Raytheon Company v. Cray, Inc.

Filing 263

MEMORANDUM AND ORDER ON CLAIM CONSTRUCTION. Signed by Magistrate Judge Roy S. Payne on 06/09/2017. (nkl, )

IN THE UNITED STATES DISTRICT COURT
FOR THE EASTERN DISTRICT OF TEXAS
MARSHALL DIVISION
RAYTHEON COMPANY,
Plaintiff,
v.
CRAY, INC.,
Defendant.

§
§
§
§
§
§
§
§
§

NO. 2:15-CV-1554-JRG-RSP

MEMORANDUM AND ORDER ON CLAIM CONSTRUCTION

On August 4, 2016, the Court held a hearing to determine the proper construction
of disputed claim terms in United States Patents No. 7,475,274, 8,190,714, 8,335,909,
and 9,037,833. Having reviewed the arguments made by the parties at the hearing and in
their claim construction briefing (Dkt. Nos. 57, 67, and 71),1 having considered the intrinsic evidence, and having made subsidiary factual findings about the extrinsic evidence, the Court issues this Claim Construction Memorandum and Order. See Phillips v.
AWH Corp., 415 F.3d 1303, 1314 (Fed. Cir. 2005); Teva Pharm. USA, Inc. v. Sandoz,
Inc., 135 S. Ct. 831, 841 (2015).

1

Citations to documents (such as the parties’ briefs and exhibits) in this Claim Construction Memorandum and Order refer to the page numbers of the original documents rather
than the page numbers assigned by the Court’s electronic docket unless otherwise indicated.

1 / 71

Table of Contents
I. BACKGROUND ............................................................................................................ 4 
II. LEGAL PRINCIPLES ................................................................................................. 5 
III. THE PARTIES’ STIPULATED TERMS .................................................................. 8 
IV. CONSTRUCTION OF DISPUTED TERMS ............................................................ 9 
A. “nodes” ........................................................................................................................ 9 
B. “integrated onto,” “integrating onto,” and “integrated to” ....................................... 10 
C. “associated with” ....................................................................................................... 15 
D. “configured to communicate with each other via a direct link between them” ........ 18 
E. “the switches for each of the n>8 interconnected nodes are configured in a
routerless manner such that each of the n>8 interconnected nodes has its own
switch” ...................................................................................................................... 19 
F. “Northbridge” ............................................................................................................ 23 
G. “topology enabled by the first switch of each of the plurality of nodes” ................. 24 
H. “folded topology”...................................................................................................... 25 
I. “distributed application”............................................................................................. 26 
J. “switching fabric” ...................................................................................................... 30 
K. “host” ........................................................................................................................ 33 
L. “boot image”.............................................................................................................. 37 
M. “[resetting / reset] a boot image of the selected node” ............................................ 38 
N. “policy associated with the distributed application” ................................................ 41 
O. “dynamically [selecting / select] one of a plurality of nodes” .................................. 45 

2 / 71

P. “[associating / associate] a virtual disk image with the selected node” .................... 49 
Q. “low utilization” ........................................................................................................ 50 
R. “the distributed application is operable to execute at a subset of the plurality of
nodes” ....................................................................................................................... 53 
S. “a manager coupled to a fabric, the manager operable to monitor a currently
running node in the HPC system executing a host and, if a fault occurs at the
currently running node, discontinue operation of the currently running node and
boot the host at a free node in the HPC system from the storage” .......................... 54 
T. “means for monitoring a currently running node in an HPC system comprising a
plurality of nodes” .................................................................................................... 60 
U. “means for, if a fault occurs at the currently running node: discontinuing
operation of the currently running node; and booting the host at a free node in
the HPC system from the storage” ........................................................................... 66 
V. CONCLUSION ............................................................................................................ 69 
APPENDIX A ................................................................................................................... 70 

3 / 71

I. BACKGROUND
Plaintiff has alleged infringement of United States Patents No. 7,475,274 (“the
’274 Patent”), 8,190,714 (“the ’714 Patent”), 8,335,909 (“the ’909 Patent”), and
9,037,833 (“the ’833 Patent”), which Plaintiff submits “relate to high performance computing (‘HPC’) technology, i.e., supercomputers.” [Dkt. # 57] at 1.
The ’274 Patent, titled “Fault Tolerance and Recovery in a High-Performance
Computing (HPC) System,” issued on January 6, 2009, and bears a filing date of November 17, 2004. The Abstract of the ’274 Patent states:
In one embodiment, a method for fault tolerance and recovery in a highperformance computing (HPC) system includes monitoring a currently running node in an HPC system including multiple nodes. A fabric coupling
the multiple nodes to each other and coupling the multiple nodes to storage
accessible to each of the multiple nodes and capable of storing multiple
hosts that are each executable at any of the multiple nodes. The method includes, if a fault occurs at the currently running node, discontinuing operation of the currently running node and booting the host at a free node in the
HPC system from the storage.
The ’714 Patent, titled “System and Method for Computer Cluster Virtualization
Using Dynamic Boot Images and Virtual Disk,” issued on May 29, 2012, and bears a filing date of April 15, 2004. The Abstract of the ’714 Patent states:
A method for computer cluster virtualization includes selecting a distributed application. A policy associated with the distributed application is retrieved. One of a plurality of nodes is dynamically selected. Then, a boot
image of the selected node is reset based, at least in part, on the retrieved
policy, with the boot image being compatible with the distributed application. Then, a virtual disk image is associated with the node. At least a portion of the distributed application is then executed on the reset node using
the associated virtual disk image.

4 / 71

The ’909 Patent, titled “Coupling Processors to Each Other for High Performance
Computing (HPC),” issued on December 18, 2012, and bears a filing date of April 15,
2004. The ’833 Patent is a continuation of the ’909 Patent, and “[a]s such, the ’909 and
’833 patents share a specification.” [Dkt. 67] at 2. The Abstract of the ’909 Patent states:
A High Performance Computing (HPC) node comprises a motherboard, a
switch comprising eight or more ports integrated on the motherboard, and
at least two processors operable to execute an HPC job, with each processor
communicably coupled to the integrated switch and integrated on the motherboard.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with preliminary constructions with the aim of focusing the parties’ arguments and
facilitating discussion. Those preliminary constructions are set forth below within the
discussion for each term.
II. LEGAL PRINCIPLES
“It is a ‘bedrock principle’ of patent law that ‘the claims of a patent define the invention to which the patentee is entitled the right to exclude.’” Phillips v. AWH Corp.,
415 F.3d 1303, 1312 (Fed. Cir. 2005) (en banc) (quoting Innova/Pure Water Inc. v. Safari
Water Filtration Sys., Inc., 381 F.3d 1111, 1115 (Fed. Cir. 2004)). Claim construction is
clearly an issue of law for the court to decide. Markman v. Westview Instruments, Inc., 52
F.3d 967, 970–71 (Fed. Cir. 1995) (en banc), aff’d, 517 U.S. 370 (1996). “In some cases,
however, the district court will need to look beyond the patent’s intrinsic evidence and to
consult extrinsic evidence in order to understand, for example, the background science or
the meaning of a term in the relevant art during the relevant time period.” Teva Pharms.

5 / 71

USA, Inc. v. Sandoz, Inc., 135 S. Ct. 831, 841 (2015) (citation omitted). “In cases where
those subsidiary facts are in dispute, courts will need to make subsidiary factual findings
about that extrinsic evidence. These are the ‘evidentiary underpinnings’ of claim construction that we discussed in Markman, and this subsidiary factfinding must be reviewed
for clear error on appeal.” Id. (citing 517 U.S. 370).
To determine the meaning of the claims, courts start by considering the intrinsic
evidence. See Phillips, 415 F.3d at 1313; see also C.R. Bard, Inc. v. U.S. Surgical Corp.,
388 F.3d 858, 861 (Fed. Cir. 2004); Bell Atl. Network Servs., Inc. v. Covad Commc’ns
Group, Inc., 262 F.3d 1258, 1267 (Fed. Cir. 2001). The intrinsic evidence includes the
claims themselves, the specification, and the prosecution history. See Phillips, 415 F.3d at
1314; C.R. Bard, 388 F.3d at 861. Courts give claim terms their ordinary and accustomed
meaning as understood by one of ordinary skill in the art at the time of the invention in
the context of the entire patent. Phillips, 415 F.3d at 1312–13; accord Alloc, Inc. v. Int’l
Trade Comm’n, 342 F.3d 1361, 1368 (Fed. Cir. 2003).
The claims themselves provide substantial guidance in determining the meaning of
particular claim terms. Phillips, 415 F.3d at 1314. First, a term’s context in the asserted
claim can be very instructive. Id. Other asserted or unasserted claims can aid in determining the claim’s meaning because claim terms are typically used consistently throughout
the patent. Id. Differences among the claim terms can also assist in understanding a
term’s meaning. Id. For example, when a dependent claim adds a limitation to an inde-

6 / 71

pendent claim, it is presumed that the independent claim does not include the limitation.
Id. at 1314–15.
“[C]laims ‘must be read in view of the specification, of which they are a part.’” Id.
at 1315 (quoting Markman, 52 F.3d at 979 (en banc)). “[T]he specification ‘is always
highly relevant to the claim construction analysis. Usually, it is dispositive; it is the single
best guide to the meaning of a disputed term.’” Phillips, 415 F.3d at 1315 (quoting
Vitronics Corp. v. Conceptronic, Inc., 90 F.3d 1576, 1582 (Fed. Cir. 1996)); accord Teleflex, Inc. v. Ficosa N. Am. Corp., 299 F.3d 1313, 1325 (Fed. Cir. 2002). This is true because a patentee may define his own terms, give a claim term a different meaning than
the term would otherwise possess, or disclaim or disavow the claim scope. Phillips, 415
F.3d at 1316. In these situations, the inventor’s lexicography governs. Id. The specification may also resolve the meaning of ambiguous claim terms “where the ordinary and accustomed meaning of the words used in the claims lack sufficient clarity to permit the
scope of the claim to be ascertained from the words alone.” Teleflex, 299 F.3d at 1325.
But, “[a]lthough the specification may aid the court in interpreting the meaning of disputed claim language, particular embodiments and examples appearing in the specification
will not generally be read into the claims.” Comark Commc’ns, Inc. v. Harris Corp., 156
F.3d 1182, 1187 (Fed. Cir. 1998) (quoting Constant v. Advanced Micro-Devices, Inc., 848
F.2d 1560, 1571 (Fed. Cir. 1988)); accord Phillips, 415 F.3d at 1323.
The prosecution history is another tool to supply the proper context for claim construction because a patent applicant may also define a term in prosecuting the patent.

7 / 71

Home Diagnostics, Inc., v. Lifescan, Inc., 381 F.3d 1352, 1356 (Fed. Cir. 2004) (“As in
the case of the specification, a patent applicant may define a term in prosecuting a patent.”). “[T]he prosecution history (or file wrapper) limits the interpretation of claims so
as to exclude any interpretation that may have been disclaimed or disavowed during
prosecution in order to obtain claim allowance.” Standard Oil Co. v. Am. Cyanamid Co.,
774 F.2d 448, 452 (Fed. Cir. 1985).
Although extrinsic evidence can be useful, it is “less significant than the intrinsic
record in determining the legally operative meaning of claim language.” Phillips, 415
F.3d at 1317 (citations and internal quotation marks omitted). Technical dictionaries and
treatises may help a court understand the underlying technology and the manner in which
one skilled in the art might use claim terms, but technical dictionaries and treatises may
provide definitions that are too broad or may not be indicative of how the term is used in
the patent. Id. at 1318. Similarly, expert testimony may aid a court in understanding the
underlying technology and determining the particular meaning of a term in the pertinent
field, but an expert’s conclusory, unsupported assertions as to a term’s definition are entirely unhelpful to a court. Id. Generally, extrinsic evidence is “less reliable than the patent and its prosecution history in determining how to read claim terms.” Id.
III. THE PARTIES’ STIPULATED TERMS
The parties reached agreement on constructions as stated in their May 19, 2016
Joint Claim Construction and Prehearing Statement. [Dkt. # 55] at 2–3. Those agreements
are set forth in Appendix A to the present Claim Construction Memorandum and Order.

8 / 71

IV. CONSTRUCTION OF DISPUTED TERMS
A. “nodes”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.

“a physical processing device with a unique
If the Court determines a construction is
address that identifies that device to all othrequired, a person having ordinary skill in
ers on the network”
the art (“PHOSITA”) would understand this
term to mean, “computing devices.”
[Dkt. # 55] at 3, 5; [Dkt. # 57] at 4; [Dkt. # 67] at 3; [Dkt. # 71] at 1. The parties submit
that this disputed term appears in Claims 1, 4, 9, 12–14, and 23–32 of the ’909 Patent,
Claims 1, 8, 11, 15, and 18–19 of the ’833 Patent, Claims 1–3, 5–10, 12–18, 20–25, 27–
29, 31–36, and 38 of the ’274 Patent, and Claims 1–9, 11–23, 25–37, and 39–42 of the
’714 Patent. [Dkt. # 55] at 3, 5.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “computing devices.” At the August 4,
2016 hearing, the parties agreed to the Court’s preliminary construction.
The Court therefore construes “nodes” to mean “computing devices.”

9 / 71

B. “integrated onto,” “integrating onto,” and “integrated to”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “mechanically and electrically connect[ed/ing] to/onto.”

“fabricated on the same motherboard without the use of daughterboards or other
components”

[Dkt. # 55] at 3–4, 6; [Dkt. # 57] at 6; [Dkt. # 67] at 5; [Dkt. # 71] at 3. The parties submit that these disputed terms appear in Claims 1, 15, and 23 of the ’909 Patent and
Claims 1, 16, 27, and 38 of the ’274 Patent. [Dkt. # 55] at 3–4, 6.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (reject defendant’s proposed ‘same motherboard’ and ‘without the use of daughterboards’ limitations).”
(1) The Parties’ Positions
Plaintiff submits that Defendant’s proposed construction is contrary to the disclosure of daughterboards, and Plaintiff argues that Defendant cannot show any prosecution
disclaimer. [Dkt. # 57] at 7–8; see id. at 19. Plaintiff also highlights that Figure 3B and
the accompanying description disclose a switch that is integrated to, but not within, the
motherboard. Id. at 7.
Defendant responds:
The ’909 and ’274 patents claim nodes comprising at least two processors
and a switch ‘integrated’ onto a motherboard. E.g., ’909 patent claim 1;

10 / 71

’274 patent claim 1. The specifications repeatedly emphasize that these
components are integrated, or fabricated, onto the motherboard as opposed
to being located externally to the motherboard. E.g., ’274 patent at 4:64–67
and ’909 patent at 4:10–13 (integrated switch minimizes distances between
nodes); ’909 patent at 7:57–62 (integration eliminates centralized functionality).
[Dkt. # 67] at 5. Defendant also cites prosecution history. Id. at 5–6. Finally, Defendant
argues that Plaintiff’s proposal of “mechanically and electrically connected to” has no
support in the intrinsic evidence. Id. at 6.
Plaintiff replies that “fabricated” “is not defined or found anywhere in the patents,” and Plaintiff reiterates that the specification discloses the use of daughterboards
integrated to the motherboard of an HPC system. [Dkt. # 71] at 3.
At the August 4, 2016 hearing, Defendant urged that the patentee’s arguments during prosecution foreclose whatever additional scope the patentee may have originally attempted to encompass with Figure 3B.
(2) Analysis
Claim 1 of the ’274 Patent, for example, recites (emphasis added):
1. A system for fault tolerance and recovery in a high-performance computing (HPC) system, the system for fault tolerance and recovery comprising:
a fabric coupling a plurality of nodes in an HPC system to each other, each node comprising a switching fabric integrated to a card
and at least two processors integrated to the card;
storage coupled to the fabric and accessible to each of the nodes, the
storage operable to store a plurality of hosts each executable at
any of the nodes; and
a manager coupled to the fabric, the manager operable to monitor a
currently running node in the HPC system executing a host and,

11 / 71

if a fault occurs at the currently running node, discontinue operation of the currently running node and boot the host at a free node
in the HPC system from the storage.
The ’909 Patent likewise discloses that a switch can be “integrated” with a blade,
and these disclosures include a switch that is illustrated as being separate from the motherboard:
Blade 315 is an integrated fabric architecture that distributes the fabric
switching components uniformly across nodes 115 in grid 110 . . . . More
specifically, blade 315 includes an integrated switch 345.
’909 Patent at 7:57–62 (emphasis added). Figure 3B of the ’909 Patent illustrates “N-Port
Switch 345” as separate from the motherboard and is reproduced here:

12 / 71

Further, Figure 3C of the ’909 Patent illustrates a blade 315 with a “Daughter
Board,” but the specification nonetheless states that “[b]lade 315 is an integrated fabric
architecture.” ’909 Patent at 7:57 (emphasis added). Figure 3C is reproduced here:

Defendant has cited prosecution history of the ’909 Patent in which the patentee
argued that the “Osten” reference failed to disclose utilizing a single board. Dec. 5, 2006
Response [Dkt. # 67-7] at 13–14. The claims at issue, however, recited “integrated on” or
“integrated with,” not integrated “to” or integrated “onto.” Id. at 6–7, 9 & 13. Also, the
patentee’s arguments centered on components being separated by a backplane slot connector. See id. at 12–14. On balance, Defendant has failed to show that the patentee made
any definitive statement as to the disputed terms. See Omega Eng’g v. Raytek Corp., 334

13 / 71

F.3d 1314, 1324 (Fed. Cir. 2003) (“As a basic principle of claim interpretation, prosecution disclaimer promotes the public notice function of the intrinsic evidence and protects
the public’s reliance on definitive statements made during prosecution.”) (emphasis added); see also Golight, Inc. v. Wal-Mart Stores, Inc., 355 F.3d 1327, 1332 (Fed. Cir. 2004)
(“Because the statements in the prosecution history are subject to multiple reasonable interpretations, they do not constitute a clear and unmistakable departure from the ordinary
meaning of the term . . . .”).
Later, during the same prosecution, the patentee argued that in the “Karpoff” reference “the CPUs and the switch reside on opposite sides of a server boundary, which
clearly teaches away from the CPUs and the switch in Karpoff being integrated on the
same motherboard, as independent Claim 1 recites.” May 29, 2007 Response [Dkt. # 679] at 7. Here again, the claim language at issue recited integrated “on” or “with,” not integrated “to” or “onto,” see id. at 2–3 & 5, and the patentee focused on the distinction of
a server boundary. See Omega Eng’g, 334 F.3d at 1324; see also Golight, 355 F.3d at
1332.
Finally, above-reproduced Claim 1 of the ’274 Patent recites “a fabric coupling a
plurality of nodes in an HPC system to each other, each node comprising a switching fabric integrated to a card.” This recital of a fabric that connects multiple nodes but that is
nonetheless “integrated to a card” reinforces that a structure need not be fabricated as
part of whatever structure it is “integrated to.”

14 / 71

The Court therefore expressly rejects Defendant’s proposal of “fabricated on the
same motherboard without the use of daughterboards or other components.” Nonetheless,
“[t]he Court believes that some construction of the disputed claim language will assist the
jury to understand the claims.” See TQP Dev., LLC v. Merrill Lynch & Co., Inc., No.
2:08-CV-471, 2012 WL 1940849, at *2 (E.D. Tex. May 29, 2012) (Bryson, J.). As to
Plaintiff’s alternative proposal of “mechanically and electrically connect[ed/ing] to/onto,”
however, the word “connected” is potentially overbroad because it might be interpreted
as allowing for many intermediate layers of connections. Instead, the word “affixed”
more appropriately conveys the required degree of directness, particularly in light of surrounding claim language reciting that components are integrated to or onto a card or
motherboard.
The Court therefore construes “integrated onto,” “integrating onto,” and “integrated to” to mean “mechanically and electrically affix[ed/ing] [to/onto].”
C. “associated with”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “related with.”

“fabricated on the same motherboard without the use of daughterboards or other
components”

[Dkt. # 55] at 4; [Dkt. # 57] at 9; [Dkt. # 67] at 7; [Dkt. # 71] at 3. The parties submit that
this disputed term appears in Claims 1 and 8 of the ’833 Patent. [Dkt. # 55] at 4.

15 / 71

Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (reject defendant’s argument that ‘associated with’ has the same meaning as ‘integrated onto’).’”
(1) The Parties’ Positions
Plaintiff argues that “[t]he use of the ‘integrated’ terms in claims of the ʼ909 patent
demonstrates that ‘associated with’ means ‘related with,’ which is a separate and distinct
meaning from the ‘integrated’ terms.” [Dkt. # 57] at 10. Plaintiff also argues that Defendant’s proposal would improperly exclude a disclosed embodiment. Id. at 10–11.
Defendant responds that “[t]he entirety of the intrinsic record makes clear that ‘associated with’ carries the same meaning as ‘integrated onto.’” [Dkt. # 67] at 7. Defendant
urges that Plaintiff’s proposal should be rejected because it is inconsistent with the prosecution history and because “the specification nowhere uses the term ‘associated with’ to
describe the relationship between processors, switches, and motherboards.” Id. at 9.
Plaintiff replies that different terms are presumed to have different meanings, and
“other than attorney argument that the (nonexistent) prosecution disclaimer of ‘integrated’ in the ’909 patent attaches to ‘associated’ in the ’833 patent, [Defendant] alleges no
other support for limiting ‘associated with’ as [Defendant] proposes.” [Dkt. # 71] at 4.
(2) Analysis
As a threshold matter, Defendant’s proposed construction should be rejected for
substantially the same reasons as discussed regarding the “integrated” terms, above, particularly as to the prosecution history cited by Defendant.

16 / 71

As to Plaintiff’s argument that “associated with” must have a meaning different
from “integrated onto,” Claim 1 of the ’833 Patent for example recites “configuring at
least two first processors to be associated with a first motherboard” and “configuring a
first switch to be associated with the first motherboard and coupling the first switch to the
first processors.” Plaintiff argues that different terms are presumed to have different
meanings, but Plaintiff has not shown any instance in which “associated with” and “integrated with” appear in claims of the same patent, let alone within the same claim.
Nonetheless, the specification uses “associated” as well as “integrated,” and Defendant has not demonstrated that the terms are used interchangeably. See, e.g., ’833 Patent at 4:33–34 (“management node 105 may periodically poll agent 132 to determine the
status of the associated node 115”), 7:25–27 (“virtual cluster 220 may be associated with
one research group, a department, a lab, or any other group of users likely to submit similar jobs 150”) & 16:53–55 (“Cluster management engine 130 then determines if a job 150
is associated with failed node 115 at decisional step 815.”); see also id. at 8:48–50 (“PCI
channel 325 comprises any high-speed, low latency link designed to increase the communication speed between integrated components.”).
The Court therefore expressly rejects Defendant’s proposal that “associated with”
has the same meaning as “integrated onto.” No further construction is necessary. See U.S.
Surgical Corp. v. Ethicon, Inc., 103 F.3d 1554, 1568 (Fed. Cir. 1997) (“Claim construction is a matter of resolution of disputed meanings and technical scope, to clarify and
when necessary to explain what the patentee covered by the claims, for use in the deter-

17 / 71

mination of infringement. It is not an obligatory exercise in redundancy.”); see also O2
Micro Int’l Ltd. v. Beyond Innovation Tech. Co., 521 F.3d 1351, 1362 (Fed. Cir. 2008)
(“[D]istrict courts are not (and should not be) required to construe every limitation present in a patent’s asserted claims.”); Finjan, Inc. v. Secure Computing Corp., 626 F.3d
1197, 1207 (Fed. Cir. 2010) (“Unlike O2 Micro, where the court failed to resolve the parties’ quarrel, the district court rejected Defendants’ construction.”); ActiveVideo Networks, Inc. v. Verizon Commcn’s, Inc., 694 F.3d 1312, 1326 (Fed. Cir. 2012); Summit 6,
LLC v. Samsung Elecs. Co., Ltd., 802 F.3d 1283, 1291 (Fed. Cir. 2015).
The Court accordingly construes “associated with” to have its plain meaning.
D. “configured to communicate with each other via a direct link between them”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “configured to communicate with each other via a point-topoint link between them.”

“capable of communicating with each other
through locally connected cache or
memory”

[Dkt. # 55] at 4; [Dkt. # 57] at 11; [Dkt. # 67] at 10. The parties submit that this disputed
term appears in Claims 1, 15, and 23 of the ’909 Patent. [Dkt. # 55] at 4.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (reject defendant’s pro-

18 / 71

posal that communication must occur ‘through locally connected cache or memory’).” At
the August 4, 2016 hearing, the parties agreed to the Court’s preliminary construction.
The Court therefore expressly rejects Defendant’s proposal that communication
must occur “through locally connected cache or memory,” and the Court construes “configured to communicate with each other via a direct link between them” to have its
plain meaning.
E. “the switches for each of the n>8 interconnected nodes are configured in a routerless manner such that each of the n>8 interconnected nodes has its own switch”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
“there is a 1-to-1 correlation between nodes
required, a PHOSITA would understand
this term to mean, “the switches for each of and switches and no routers are used to
the n>8 interconnected computing devices pass messages between nodes”
are configured such that each of the n>8
interconnected computing devices has its
own switch.”
[Dkt. # 55] at 4; [Dkt. # 57] at 12; [Dkt. # 67] at 11; [Dkt. # 71] at 4. The parties submit
that this disputed term appears in Claims 1 and 8 of the ’833 Patent. [Dkt. # 55] at 4.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (apart from the court’s
construction of the constituent term ‘nodes,’ which is addressed separately above).”

19 / 71

(1) The Parties’ Positions
Plaintiff argues that the meaning of this term is evident from surrounding claim
language. [Dkt. # 57] at 12. Plaintiff submits that Defendant’s proposal “departs from the
plain claim language” and “wholly omits the claim phrase ‘for each of the n>8 interconnected nodes.’” Id. Further, Plaintiff argues that the specification discloses each node has
an integrated switch and explains the benefits of using a node with its own switch. Id. at
12–13. Plaintiff urges that “there is no support in the specification that each switch works
exclusively with only one node, as [Defendant] proposes.” Id. at 13 (citation and alterations omitted).
Defendant responds: “By the plain claim language, ‘each’ node ‘has its own
switch.’ In other words, if there are 24 nodes, there must be 24 switches.” [Dkt. # 67] at
12. Defendant argues that this interpretation is consistent with the specification, and Defendant also cites prosecution history. Id.
Plaintiff replies that “[t]he plain language of the claims defines ‘routerless manner,’ which is reflected in the ‘such that’ portion of the claim.” [Dkt. # 71] at 4. Plaintiff
also argues that neither the specification nor the prosecution history contain any statement requiring a one-to-one correlation between nodes and switches. Id.
(2) Analysis
As Defendant has highlighted, the specification of the ’833 Patent discloses:
Blade 315 is an integrated fabric architecture that distributes the fabric
switching components uniformly across nodes 115 in grid 110, thereby pos-

20 / 71

sibly reducing or eliminating any centralized switching function, increasing
the fault tolerance, and allowing message passing in parallel.
’833 Patent at 7:63–67 (emphasis added); see id. at 8:16–24 (“Each node 115 is configured with an integrated switch.”). Also, during prosecution, the patentee distinguished the
“Thorson” reference as having switching functionality that is centralized to routers when
there are more than eight nodes. See Mar. 19, 2014 Amendment [Dkt. # 67-13] at 8–9.
Defendant’s proposed construction, however, essentially merely restates the “routerless” and “has its own switch” limitations that are already set forth in the disputed term.
At the August 4, 2016 hearing, Defendant emphasized that there must be at least as many
switches as nodes because each node must have at least one switch. Because this is already evident from the “has its own switch” language on its face, this requirement need
not be reiterated.
Finally, nothing in this claim language is contrary to allowing for a node that has
its own switch to also be connected to another switch. See, e.g., Gillette Co. v. Energizer
Holdings, Inc., 405 F.3d 1367, 1371 (Fed. Cir. 2005) (“The word ‘comprising’ transitioning from the preamble to the body signals that the entire claim is presumptively openended.”). To whatever extent Defendant has suggested to the contrary, the Court expressly rejects any such negative limitation.
The Court therefore rejects Defendant’s proposed construction. No further construction is necessary. See U.S. Surgical, 103 F.3d at 1568; see also O2 Micro, 521 F.3d at

21 / 71

1362; Finjan, 626 F.3d at 1207; ActiveVideo, 694 F.3d at 1326; Summit 6, 802 F.3d at
1291.
At the August 4, 2016 hearing, the parties also presented some disagreement as to
how to determine whether a node has “its own” switch or, in other words, whether a particular switch belongs to a particular node. Ultimately, this issue depends upon details of
particular implementations and is a question of fact regarding infringement rather than a
question of law for claim construction. See Acumed LLC v. Stryker Corp., 483 F.3d 800,
806 (Fed. Cir. 2007) (“The resolution of some line-drawing problems . . . is properly left
to the trier of fact.”) (citing PPG Indus. v. Guardian Indus. Corp., 156 F.3d 1351, 1355
(Fed. Cir. 1998) (“after the court has defined the claim with whatever specificity and precision is warranted by the language of the claim and the evidence bearing on the proper
construction, the task of determining whether the construed claim reads on the accused
product is for the finder of fact”)); see EON Corp. IP Holdings LLC v. Silver Springs
Networks, Inc., 815 F.3d 1314, 1318–19 (Fed. Cir. 2016) (citing PPG).
The Court therefore construes “the switches for each of the n>8 interconnected
nodes are configured in a routerless manner such that each of the n>8 interconnected nodes has its own switch” to have its plain meaning apart from the Court’s construction of the constituent term “nodes,” which is addressed separately above.

22 / 71

F. “Northbridge”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “a circuit that enables a
processor to communicate with memory, a
PCI bus, Level 2 cache, or any other related
components.”

“a device which communicates with the
processors and controls communications
with memory, a PCI bus, Level 2 cache, or
any other related components”

[Dkt. # 55] at 4–5; [Dkt. # 57] at 13; [Dkt. # 67] at 13. The parties submit that this disputed term appears in Claims 5 and 12 of the ’833 Patent. [Dkt. # 55] at 4–5.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “a device which communicates with the
processors and controls communications with memory, a PCI bus, level 2 cache, or any
other related components (reject defendant’s proposal of requiring a physically separate
‘device’).” At the August 4, 2016 hearing, the parties agreed to the Court’s preliminary
construction.
The Court accordingly rejects Defendant’s argument that a “Northbridge” must be
a physically separate device, and with that understanding the Court construes “Northbridge” to mean “a device which communicates with the processors and controls
communications with memory, a PCI bus, level 2 cache, or any other related components.”

23 / 71

G. “topology enabled by the first switch of each of the plurality of nodes”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
“topology enabled by the number of mothIf the Court determines a construction is
erboards to which the first switch of each
required, a PHOSITA would understand
this term to mean, “topology enabled by the motherboard can connect”
first switch of each of the two or more
nodes.”
[Dkt. # 55] at 5; [Dkt. # 57] at 14; [Dkt. # 67] at 14. The parties submit that this disputed
term appears in Claim 9 of the ’909 Patent and Claim 15 of the ’833 Patent. [Dkt. # 55] at
5.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (reject def[endant]’s
proposal of referring to motherboards).” At the August 4, 2016 hearing, the parties agreed
to the Court’s preliminary construction.
The Court accordingly rejects Defendant’s proposal of limiting this disputed term
to motherboards, and the Court construes “topology enabled by the first switch of each
of the plurality of nodes” to have its plain meaning.

24 / 71

H. “folded topology”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “topology that has links
of equal length that are minimized.”

“topology of motherboards in which the
motherboard at the physical edge of the interconnected motherboards is connected to
a corresponding motherboard at the axial
edge”

[Dkt. # 55] at 5; [Dkt. # 57] at 15; [Dkt. # 67] at 14. The parties submit that this disputed
term appears in Claim 11 of the ’909 Patent. [Dkt. # 55] at 5.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “topology of nodes in which a node at
the physical edge of the interconnected nodes is connected to a corresponding node at the
axial edge.” At the August 4, 2016 hearing, the parties agreed to the Court’s preliminary
construction.
The Court therefore construes “folded topology” to mean “topology of nodes in
which a node at the physical edge of the interconnected nodes is connected to a corresponding node at the axial edge.”

25 / 71

I. “distributed application”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “an application capable
of running across at least two nodes.”

“a single instance of an application that is
run across more than one computing device
and that appears to a client computer as a
single application”

[Dkt. # 55] at 5–6; [Dkt. # 57] at 16; [Dkt. # 67] at 15; [Dkt. # 71] at 5. The parties submit that this disputed term appears in Claims 1–2, 5–8, 11–12, 15–16, 19–22, 25–26, 29–
30, 33–34, and 39–40 of the ’714 Patent. [Dkt. # 55] at 5–6.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “a single instance of an application running across more than one node.”
(1) The Parties’ Positions
Plaintiff argues that its proposed construction is supported by the specification as
well as by the ordinary understanding of the disputed term in the relevant art. [Dkt. # 57]
at 17. Plaintiff argues that Defendant’s proposed construction is contrary to the specification, fails to cover applications that are collaborative rather than client-server, and relies
upon an inapplicable dictionary definition. Id.
Defendant responds that its proposal is consistent with the specification as well as
with the opinion of Plaintiff’s expert. [Dkt. # 67] at 16.

26 / 71

Plaintiff replies that Defendant’s proposals of “single instance” and “appears to a
client as a single application” have no support in the intrinsic evidence, and “[Defendant’s] only alleged support is the declaration of its expert and [Defendant’s] mischaracterization of an out-of-context statement by [Plaintiff’s] expert.” [Dkt. # 71] at 5.
(2) Analysis
At the August 4, 2016 hearing, Plaintiff argued that the disputed term could encompass multiple instances of the same application running on multiple nodes. Defendant
responded that Plaintiff’s position in that regard is inconsistent with the specification,
such as Figure 1 of the ’714 Patent, which illustrates applications 114a and 114b with
lines that emanate therefrom having arrows pointing to multiple nodes 108. Defendant
argued that application 114a, for example, represents a single instance of an application
and that the lines illustrate the single instance running across multiple nodes. Plaintiff replied that there is no disclosure that supports Defendant’s interpretation of Figure 1,
which Plaintiff argues could just as easily be interpreted such that each line emanating
from application 114a represents a different instance of that application. Of note in this
regard, the specification does not use reference numerals 114a and 114b but rather refers
simply to “applications 114.” See ’714 Patent at 2:16–18.
Even setting aside Figure 1, however, the specification discusses “distributed application 114” in the context of using multiple nodes:
In one aspect of operation, dynamic boot engine 105 selects a distributed
application 114. Based on one or more associated policies 132, dynamic
boot engine 105 may dynamically add or subtract one or more selected

27 / 71

nodes 108 to the particular application environment or virtual cluster.
Based on the retrieved policies 132, dynamic boot engine 105 selects the
appropriate boot image 132 for the selected nodes 108. For example, if
there are already four nodes 108 executing a portion of application 114,
then dynamic boot engine 105 automatically selects the fifth boot image
132 (at least partially based on node’s 108 hardware and other characteristics and the one or more policies 132) that is compatible with application
114. Based on the retrieved policies 132, dynamic boot engine 105 may also select the appropriate virtual local disk image 134 for the selected nodes
108. Once the appropriate boot image 132 and/or virtual local disk image
134 are selected, dynamic boot engine 105 flashes node 108 with a pointer
or other reference to the selected boot image 132 and virtual local disk image 134 and reboots node 108. Once node 108 is initialized (normally less
than fifteen seconds), dynamic boot engine 105 (or some other job scheduler) executes the appropriate task, process, or other portion of application
104 [sic, 114] on the selected node 108.
ʼ714 Patent at 7:62–8:17 (emphasis added); see also id. at 3:58–60 (“Nodes 108 comprises [sic] any computer, blade, or server operable to execute at least a portion (such as a
task or process) of application 114.”) (emphasis added) & Fig. 1.
Also, as to the term “the distributed application is operable to execute at a subset
of the plurality of nodes” (which is no longer disputed, as noted below), Defendant had
proposed that the corresponding structure is “a database management system (DBMS) or
financial software.” Plaintiff’s expert opined: “There are many more distributed applications in this world than just DBMS and financial software programs. [Defendant] limits
its construction to two types of distributed applications, but ignores that the distributed
application may be any suitable software to run a single instance on more than one node.”
Scherson Decl. (May 19, 2016) [Dkt. # 57-5] at ¶ 202 (emphasis added) (citing ’714 Pa-

28 / 71

tent 4:19–22). Although Plaintiff argues that Defendant has cited this statement out of
context,2 it is noteworthy that Plaintiff’s own expert has referred to a distributed application as being a single instance running on multiple nodes.
Defendant further submits the following extrinsic definition: “A distributed system
is a collection of independent computers that appears to its users as a single coherent system.” Andrew S. Tanenbaum, et al., Distributed Systems: Principles & Paradigms (2007)
[Dkt. # 67-24] at 2.
On balance, in light of the above-discussed intrinsic and extrinsic evidence, Defendant has adequately supported its proposal that, in the context of an application, the
term “distributed” requires that the application is a single instance running across more
than one node rather than, as Plaintiff has suggested, multiple instances of an application
being run separately by different nodes. Plaintiff’s proposal in that regard fails to give
meaning to the constituent term “distributed” and is therefore rejected.
Defendant’s proposal of “appears to a client computer as a single application,”
however, lacks adequate support, is potentially redundant, and would tend to confuse rather than clarify the scope of the claims.
The Court therefore construes “distributed application” to mean “a single instance of an application running across more than one node.”

2

See Scherson Decl. (July 14, 2016) [Dkt. # 71-1] ¶ 15 (stating that Defendant’s expert
“mischaracterizes my prior declaration because I never admit that a distributed application must run as a single instance”).

29 / 71

J. “switching fabric”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “one or more switches
on a fabric.”

“the combination of hardware and software
components, including a switch, that connects the two or more processors on the
card to the adjacent node(s) in the network”

[Dkt. # 55] at 6; [Dkt. # 57] at 17; [Dkt. # 67] at 16; [Dkt. # 71] at 5. The parties submit
that this disputed term appears in Claims 1, 16, 27, and 38 of the ’274 Patent. [Dkt. # 55]
at 6.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “a network of switches and links between switches.”
(1) The Parties’ Positions
Plaintiff argues that its proposal is consistent with surrounding claim language as
well as the specification. [Dkt. # 57] at 18.
Defendant responds that “[t]he intrinsic record demonstrates that ‘switching fabric’
is more than simply switches, as [Plaintiff] contends, but rather includes the hardware and
software connecting the processors of different nodes.” [Dkt. # 67] at 17.
Plaintiff replies that Defendant’s proposal ignores the phrase “on a fabric” and
“ignores [Plaintiff’s] explanation that the term ‘fabric’ refers to the network that allows
communications between nodes.” [Dkt. # 71] at 5. Plaintiff also argues that Defendant’s

30 / 71

proposals of “combination of hardware and software components” and “connects . . . to
the adjacent nodes” are “each unsupported by the claims and the specification.” Id. at 5–
6.
(2) Analysis
The specification discloses: “Each node 115 is configured with an integrated
switch.” ’274 Patent at 4:64–65. Also, the specification discloses: “In particular embodiments, the network fabric of HPC server 102 includes switches 166 coupled to each other
according to a topology encompassing a three dimensional torus, as described above.” Id.
at 62:46–49; see id. at Abstract (“[a] fabric coupling the multiple nodes to each other and
coupling the multiple nodes to storage accessible to each of the multiple nodes”).
At the August 4, 2016 hearing, Plaintiff argued that including the links would be
inconsistent with the recital in the claims that the “switching fabric” is “integrated to a
card” because the links cannot be built into a card. Claim 1 of the ’274 Patent, for example, recites (emphasis added):
1. A system for fault tolerance and recovery in a high-performance computing (HPC) system, the system for fault tolerance and recovery comprising:
a fabric coupling a plurality of nodes in an HPC system to each other, each node comprising a switching fabric integrated to a card
and at least two processors integrated to the card;
storage coupled to the fabric and accessible to each of the nodes, the
storage operable to store a plurality of hosts each executable at
any of the nodes; and
a manager coupled to the fabric, the manager operable to monitor a
currently running node in the HPC system executing a host and,
if a fault occurs at the currently running node, discontinue opera-

31 / 71

tion of the currently running node and boot the host at a free node
in the HPC system from the storage.
The Court separately addresses the term “integrated to,” above. Regardless of the
Court’s construction of that term, however, Plaintiff has not shown that finding that a link
is part of something that is “integrated to a card” would require that the entire link is integrated rather than merely a portion or terminus thereof. Also of note, the abovereproduced claim itself uses the term “fabric” to refer to couplings between a plurality of
nodes (rather than merely the switches at the nodes).
On balance, the above-quoted portions of the specification, together with the context of the above-reproduced claim, demonstrate that the term “switching fabric” refers to
a network of switches and links between switches. Further, this interpretation is consistent with the extrinsic definition submitted by Defendant ([Dkt. # 67-14] (“The term is
sometimes used to mean collectively all switching hardware and software in a network.”)), and is also consistent with the opinion of Plaintiff’s expert that “switching fabric” refers to switches and “a set of links” (Scherson Decl. (May 19, 2016) [Dkt. # 57-5]
¶ 143).
Finally, Defendant’s proposed reference to “processors on the card” is rejected as
tending to confuse rather than clarify and as being redundant of other claim language requiring a “card.”
The Court therefore construes “switching fabric” to mean “a network of switches and links between switches.”

32 / 71

K. “host”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “program.”

“a physical or virtual computing system
capable of providing one or more services
over a network or fabric”

[Dkt. # 55] at 6; [Dkt. # 57] at 19; [Dkt. # 67] at 18; [Dkt. # 71] at 6. The parties submit
that this disputed term appears in Claims 1, 6–7, 12, 16, 21–22, 27, 32–33, and 38 of the
’274 Patent. [Dkt.#. 55] at 6.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “operating system.”
(1) The Parties’ Positions
Plaintiff argues that whereas its proposed construction is consistent with the specification, Defendant’s proposal “imposes unnecessary, unsupported limitations. In particular, defining ‘hosts’ as only ‘services’ contradicts the specification which describes ‘hosts’
as ‘a thread or process.’” [Dkt. # 57] at 19 (citing ʼ274 Patent at 7:57–60).
Defendant responds that “[a]lthough ‘host’ is not defined in the ’274 patent, [Defendant’s] proposal captures the disclosures in the intrinsic record.” [Dkt. # 67] at 18. Defendant also argues that Plaintiff’s proposed construction “is contrary to the patent’s disclosures” because “[t]he specification never equates hosts with applications/programs; in
fact, it refers to them separately.” Id. at 19 (citing ’274 Patent at 62:40–42).

33 / 71

Plaintiff argues that “there is no disclosure of a ‘host’ being ‘physical’ or being a
‘computing system’ anywhere in the specification or claims,” and “limiting ‘hosts’ to only ‘services’ is contrary to the specification, which teaches that hosts could also refer to
the ‘processes’ of a distributed application.” [Dkt. # 71] at 6 (citing ’274 Patent at 7:57–
60).
At the August 4, 2016 hearing, the parties’ dispute crystallized into whether a
“host” must include an operating system. Defendant was amenable to the Court’s preliminary construction of “host” as meaning “operating system,” but Defendant argued that
the preliminary construction was incomplete because additional structures are required
for running an operating system. Nonetheless, Defendant stated that construing “host” as
“operating system” would adequately resolve the parties’ dispute in the present case. Defendant also noted that the parties agreed to the Court’s preliminary construction of “boot
image” (addressed below) as “operating system data that is used to initialize a node.”
Plaintiff responded that a “host” need not include an operating system. Plaintiff
emphasized dependent Claim 12 of the ’274 Patent, which recites limitations as to what a
“host” includes but which does not recite an operating system. In response to Defendant’s
arguments, however, Plaintiff stated that it would be amenable to construing “host” to
mean “program that creates an execution environment.”
(2) Analysis
Claims 1 and 12 of the ’274 Patent, for example, recite (emphasis added):

34 / 71

1. A system for fault tolerance and recovery in a high-performance computing (HPC) system, the system for fault tolerance and recovery comprising:
a fabric coupling a plurality of nodes in an HPC system to each other, each node comprising a switching fabric integrated to a card
and at least two processors integrated to the card;
storage coupled to the fabric and accessible to each of the nodes, the
storage operable to store a plurality of hosts each executable at
any of the nodes; and
a manager coupled to the fabric, the manager operable to monitor a
currently running node in the HPC system executing a host and,
if a fault occurs at the currently running node, discontinue operation of the currently running node and boot the host at a free node
in the HPC system from the storage.
12. The system of claim 1, wherein a host comprises an Internet Protocol
(IP) address, a boot image, a configuration, and a file system usable to boot
the host at a node in the HPC system.
The specification appears to distinguish between “hosts” and “applications.” See,
e.g., ’274 Patent at 62:40–42 (“Centralized storage 540 includes application data including data on hosts and applications executable at nodes 115 in grid 110, as described below.”) (emphasis added) & 65:7–10 (“restart the host and the application at a second node
115 in grid 110”).
The specification further discloses that hosts can be “instantiated”: “Instantiation
manager 534 may interact with one or more components of cluster management engine
130 (such as, for example, physical manager 505, virtual manager 510, or both) to dynamically instantiate one or more hosts at one or more nodes 115 in response to a con-

35 / 71

nection request from a client 120, according to particular needs.” Id. at 61:19–24 (emphasis added).
In particular embodiments, instantiation data 536 includes one or more boot
images for instantiating hosts at nodes 115 to provide services. In particular
embodiments, instantiation data 536 also includes one or more file systems
for instantiating hosts at nodes 115 to provide services. In particular embodiments, instantiation data 536 also includes one or more OS configuration
files for instantiating hosts at nodes 115 to provide services.
Id. at 61:60–67 (emphasis added).
As to Defendant’s proposal of “providing one or more services,” the specification
discloses that a node may perform merely a portion of a task. See ’274 Patent at 7:58–61
(“A node 115 includes any computing device in any orientation for processing all or a
portion, such as a thread or process, of one or more jobs 150.”).
As to Defendant’s proposal of a “computing system,” the above-quoted claim refers to booting a host on a node or at a node. The specification is similar. See id. at 2:10,
63:9–21, 64:20–21, 68:64–65 & Abstract. The specification is also consistent with Defendant’s proposal of something more than merely a program. See, e.g., ’274 Patent at
61:25–62:17. Also, the opinion of Defendant’s expert that “one would not use the verb
‘boot’ to refer to software” is persuasive to the extent that what is booted is not merely a
program but rather is something capable of running a program. See Livny Decl. (June 30,
2016) [Dkt. # 67-3] ¶ 25 n.1.
As to extrinsic evidence, Defendant has submitted dictionary definitions of “host”
as meaning: “An intelligent device attached to a network” (Newton’s Telecom Dictionary

36 / 71

(16th ed. 2000) [Dkt. # 67-18] (CRAY00021621)); and “A computer containing data or
programs that another computer can access over a network or by modem” (Dictionary of
Computer Words 133 (1995) [Dkt. # 67-19]). These definitions thus reinforce Defendant’s
proposal that a “host” is not merely a program but rather is something that can execute a
program.
On balance, Defendant has persuasively shown that a “host” is an environment in
which programs can be executed, and therefore a host includes an operating system. See,
e.g., ’274 Patent at 7:34 (“Cluster management engine 130 may also manage and provide
a unique execution environment for each allocated node of virtual cluster 220.”). Indeed,
at the August 4, 2016 hearing, Plaintiff alternatively proposed that a “host” is a “program
that creates an execution environment.”
The Court therefore construes “host” to mean “execution environment, including an operating system.”
L. “boot image”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “boot file.”

“any form, image, pointer, or reference to
at least a portion of the boot drive primary
operating system partition”

37 / 71

[Dkt. # 55] at 6; [Dkt. # 57] at 19; [Dkt. # 67] at 20; [Dkt. # 71] at 6. The parties submit
that this disputed term appears in Claim 12 of the ’274 Patent and Claims 1, 2, 4, 8, 10–
11, 15–18, 22, 24, 29–32, 36, and 38–39 of the ’714 Patent. [Dkt. # 55] at 6.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “operating system data that is used to
initialize a node.” At the August 4, 2016 hearing, the parties agreed to the Court’s preliminary construction.
The Court therefore construes “boot image” to mean “operating system data
that is used to initialize a node.”
M. “[resetting / reset] a boot image of the selected node”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.

“[changing/change] the selected node’s
form, image, pointer, or reference to at least
If the Court determines a construction is
a portion of a different boot drive primary
required, a PHOSITA would understand
operating system partition from its current
this term to mean, “[resetting / reset] a boot setting”
image of the selected node.”
[Dkt. # 55] at 7; [Dkt. # 57] at 23; [Dkt. # 67] at 21; [Dkt. # 71] at 8. The parties submit
that these disputed terms appear in Claims 1, 15, and 29 of the ’714 Patent. [Dkt. # 55] at
7.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (reject def[endant]’s
proposal of ‘changing’).”

38 / 71

(1) The Parties’ Positions
Plaintiff argues that whereas its proposed construction “is consistent with the understanding of a POSITA [(person of ordinary skill in the art at the time of the invention)]
in view of the specification,” “[Defendant] improperly limits resetting a boot image to
changing the selected node’s form, image, pointer, or reference, a proposed construction
improperly imported from the specification.” [Dkt. # 57] at 23.
Defendant responds that because “reset” is used in the specification in two different contexts, namely resetting a node and resetting a boot image, “the term ‘reset’ as used
in the patent lacks clarity, and should be construed in the context of its use in the disputed
claim term.” [Dkt. # 67] at 21. Defendant argues that “the ‘resetting’ step does not merely
mean that a boot image is reapplied to a node; the node must be reset using a different
boot image.” Id. at 22.
Plaintiff replies that “the specification passage [Defendant] cites . . . indicates that
a different boot image is possible, but not required.” [Dkt. # 71] at 8 (citing ’714 Patent at
3:31–35).
(2) Analysis
The specification discloses embodiments in which resetting a node involves selecting a different boot image:
Generally, dynamic boot engine 105 manages one or more applications 114
by starting and stopping application environments on the individual nodes
108. For example, dynamic boot engine 105 may reset the particular node
108 with a different boot image 131 from boot image file 130, which is specific to or compatible with the desired application environment.

39 / 71

’714 Patent at 3:29–35 (emphasis added).
If node 108 is not compatible with application 114, then dynamic boot engine 105 brings down the selected node using any suitable technique at step
225. . . . Based on the selected policy 132, dynamic boot engine 105 flashes
the selected node with a pointer to a new boot image 131 at step 235 and
associates virtual local disk image 134 at step 237. As described above, dynamic boot engine 105 may flash EEPROM 111 or any other suitable component. Next, dynamic boot engine 105 boots the selected node 108 using
the new boot image 131 at step 240.
Id. at 8:51–66 (emphasis added); see id. at 8:25–28.
The Court has addressed the constituent term “boot image” separately, above. As
to Defendant’s argument that “resetting” means that a different boot image must be used,
Defendant has not adequately justified such a limitation. No such lexicography or disclaimer is evident. Although Defendant has argued that the context of surrounding claim
demonstrates that “resetting” occurs only if a different boot image is needed, no such limitation is apparent in the claims at issue. See ’714 Patent at Claims 1, 15 & 29.
Instead, this is a specific feature of particular embodiments that should not be imported into the claims. See Comark, 156 F.3d at 1187; see also Phillips, 415 F.3d at 1323.
Also, analogously, the specification discloses “recover[ing] server 102 from single hardware component failures by automatically replacing and dynamically rebooting a replacement node 108 for the failed node 108.” ’714 Patent at 3:51–54 (emphasis added).
From this disclosure can be inferred the potential desirability of restarting a failed node
using the same boot image rather than a different boot image.

40 / 71

The Court therefore expressly rejects Defendant’s proposed construction. No further construction is necessary. See U.S. Surgical, 103 F.3d at 1568; see also O2 Micro,
521 F.3d at 1362; Finjan, 626 F.3d at 1207; ActiveVideo, 694 F.3d at 1326; Summit 6, 802
F.3d at 1291.
The Court therefore construes “[resetting / reset] a boot image of the selected
node” to have its plain meaning.
N. “policy associated with the distributed application”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “requirements for the
distributed application.”

“set of variables that define the required
characteristics of a node in order to run the
distributed application, including at least
information related to a boot image and a
virtual disk image”

[Dkt. # 55] at 7; [Dkt. # 57] at 21; [Dkt. # 67] at 22; [Dkt. # 71 at 7. The parties submit
that this disputed term appears in Claims 1, 15, and 29 of the ’714 Patent. [Dkt. # 55] at
7.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “one or more parameters that define the
required characteristics of a node in order to run the distributed application.”
(1) The Parties’ Positions
Plaintiff argues that whereas its proposed construction is consistent with the specification, “[Defendant’s] construction improperly ties ‘policy’ to ‘the required characteris-

41 / 71

tics of a node’ rather than a ‘distributed application’” and “is overly-narrow because it
limits ‘set of variables’ to ‘information related to a boot image and a virtual disk image.’”
[Dkt. # 57] at 21.
Defendant responds that their proposed construction “takes into account the context of the claims, which reflects that the policy contains information concerning the boot
image and virtual disk.” [Dkt. # 67] at 22. Defendant also cites prosecution history. Id. at
22–23. Further, Defendant submits that its proposed construction “does not limit the contents of the policy to any particular information; rather, it includes any content whatsoever as long as it includes the two types of information the inventors considered mandatory
given the focus of the claims.” Id. at 23.
Plaintiff replies that “[Defendant] fails to explain why its proposal is more helpful
[than] just using the word ‘requirements,’” and “[t]he prosecution history also does not
support [Defendant’s] proposal.” [Dkt. # 71] at 7.
(2) Analysis
Claim 1 of the ’714 Patent, for example, recites (emphasis added):
1. A method comprising:
selecting a distributed application;
retrieving a policy associated with the distributed application;
dynamically selecting one of a plurality of nodes;
resetting a boot image of the selected node based at least in part on
the retrieved policy, wherein the boot image being compatible
with the distributed application, wherein resetting the boot image
of the selected node comprises: automatically shutting down the

42 / 71

selected node; resetting the boot image of the selected node; and
restarting the selected node using the reset boot image;
associating a virtual disk image with the selected node based at least
in part on the retrieved policy; and
executing at least a portion of the distributed application on the selected node, as reset, using the virtual disk image associated with
the selected node, the execution performed by at least one processor of the selected node.
Surrounding claim language thus provides that the “policy associated with the distributed application” pertains to selecting an appropriate boot image and virtual disk image based on characteristics that are necessary for running the distributed application.
Such limitations, however, are set forth by the surrounding claim language. Defendant
has not demonstrated that such limitations are inherent features of the “policy.”
The specification discloses:
Policies table 132 includes any parameters for managing nodes 108 and applications 114. For example, policies 132 may be for automatically adding
or subtracting nodes 108 to application environments.
’714 Patent at 5:45–48; see id. at 5:45–6:18. The specification thus better supports referring to “parameters” rather than Defendant’s proposal of “variables.”
Defendant has also cited prosecution history in which the patentee argued:
Furthermore, the cited portions of Lortz fail to disclose . . . that the policy
in Lortz has anything to do with a boot image or a virtual disk image such
that a boot image of a selected node could be reset based at least in part of
[sic, on] the retrieved policy and that a virtual disk image could be associated with the selected node based at least in part of [sic, on] the retrieved policy, as would be required for the [sic] Lortz to even possibly disclose the retrieved policy of Claim 1.

43 / 71

Jan. 27, 2011 Response [Dkt. # 67-20] at 13; see July 15, 2011 Pre-Appeal Brief Request
for Review [Dkt. # 67-21] at 2 (similar); see also Sept. 26, 2011 Appeal Brief [Dkt. # 6722] at 19–20 (similar).
On balance, this passage does not give rise to a clear disclaimer so as to require
that the policy must include information related to a boot image and a virtual disk image.
See Omega Eng’g, 334 F.3d at 1324 (“As a basic principle of claim interpretation, prosecution disclaimer promotes the public notice function of the intrinsic evidence and protects the public’s reliance on definitive statements made during prosecution.”) (emphasis
added); see also Golight, 355 F.3d at 1332 (“Because the statements in the prosecution
history are subject to multiple reasonable interpretations, they do not constitute a clear
and unmistakable departure from the ordinary meaning of the term . . . .”). In particular,
Defendant has not shown that selecting a boot image or a virtual disk image based on a
policy necessarily means that the policy itself contains information about a boot image or
a virtual disk image. Instead, this prosecution history merely sets forth limitations that
appear in surrounding claim language.
Finally, whereas Defendant has proposed construing this disputed term with reference to “characteristics of a node,” the specification discloses that an application may
need multiple nodes. See ’714 Patent at 5:45–48 (quoted above); see also id. at 5:61–63
(“policies table 132 may store individual virtual cluster policies including:
i) minimum/maximum number of nodes 108 assigned to an application environment;
. . . .”). The construction should therefore refer instead to an “execution environment.”

44 / 71

The Court rejects Plaintiff’s argument that the policy refers to characteristics of an application because, in the above-cited prosecution history, the patentee distinguished the policy in the Lortz reference as not having “anything to do with a boot image or a virtual
disk image.” Jan. 27, 2011 Response [Dkt. # 67-20] at 13; see Typhoon Touch Techs., Inc.
v. Dell, Inc., 659 F.3d 1376, 1381 (Fed. Cir. 2011) (“The patentee is bound by representations made and actions that were taken in order to obtain the patent.”).
The Court therefore construes “policy associated with the distributed application” to mean “one or more parameters that define the required characteristics of an
execution environment in order to run the distributed application.”
O. “dynamically [selecting / select] one of a plurality of nodes”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “[selecting/select] a
node, at-least-in-part at run-time, based on
one or more variables.”

“[selecting/select] a node at least in part at
run-time based on one or more variables,
excluding manually selecting nodes”

[Dkt. # 55] at 7; [Dkt. # 57] at 22; [Dkt. # 67] at 23; [Dkt. # 71] at 7. The parties submit
that these disputed terms appear in Claims 1, 5, 15, 19, 29, and 33 of the ’714 Patent.
[Dkt. # 55] at 7.

45 / 71

Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “[selecting/select] a node at least in part
at run-time based on one or more variables.”
(1) The Parties’ Positions
Plaintiff argues that “‘[d]ynamically’ is explicitly defined by the patent.” [Dkt. #
57] at 22 (citing ’714 Patent at 2:33–35, 3:10–14, 3:49–57, 7:34–49 & 7:62–66). Plaintiff
submits that Defendant’s proposal is inconsistent with that express definition, and Plaintiff argues that “by ‘excluding manually selecting nodes,’ [Defendant] improperly reads
out an embodiment allowing for manually selecting nodes on-the-fly.” [Dkt. # 57] at 22.
Defendant responds that the specification and the prosecution history distinguish
between dynamic selection and manual selection. [Dkt. # 67] at 23–24. Defendant also
argues that Plaintiff has failed to demonstrate that Defendant’s proposal reads out any
disclosed embodiments. Id. at 24.
Plaintiff replies that “[Defendant’s] proposed limitation finds no support in the
specification or any ‘clear and unmistakable’ disavowal in the prosecution history.” [Dkt.
# 71] at 8.
(2) Analysis
The parties appear to agree that the following passage sets forth a definition for
“dynamically” or, at least, that the following specified attributes of “dynamically” should
be included in the construction of the disputed term:

46 / 71

In short, system 100 is any computing environment that automatically allows nodes 108 to be dynamically allocated on-the-fly as application 114
requirements, parameters, and processing needs change. The term “dynamically,” as used herein, generally means that certain processing is determined, at least in part, at run-time based on one or more variables.
’714 Patent at 2:29–35.
The remaining dispute is whether “dynamically” excludes “manually.” On one
hand, the specification contrasts the terms “dynamically” and “manually” as follows:
Method 200 begins at step 205, where dynamic boot engine 105 determines
that software application 114 should be allocated more nodes 108. This determination may occur using any appropriate technique. For example, the
administrator may manually add node 108 to the application environment
for application 114. In another example, dynamic boot engine 105 may dynamically determine that nodes 108 may or should be used based on policies 132.
Id. at 8:34–41 (emphasis added).
On the other hand, the specification discloses that an administrator may make
changes:
In one embodiment, GUI 122 may allow an administrator to create, delete,
copy, and modify application environments. The administrator may also set
up application environment sharing policies, activate and deactivate application environments, monitor states and loads of application environments
and nodes 108 using GUI 122. Further, GUI 122 may allow the adding or
subtracting of nodes 108 from active application environments.
Id. at 7:34–41 (emphasis added). Moreover, the term “automatically” is specifically described apart from “dynamically,” which demonstrates that the ’714 Patent does not use
“dynamically” to mean “automatically”:

47 / 71

The term “dynamically,” as used herein, generally means that certain processing is determined, at least in part, at run-time based on one or more variables. The term “automatically,” as used herein, generally means that the
appropriate processing is substantially performed by at least part of system
100. It should be understood that “automatically” further contemplates any
suitable user or administrator interaction with system 100 without departing
from the scope of this disclosure.
Id. at 2:33–41. In addition, this passage further counsels against Defendant’s proposed
negative limitation by noting that even the term “automatically” can encompass some
amount of “user or administrator interaction.” Id. The intrinsic evidence thus demonstrates that, in the ’714 Patent, “dynamically” merely means at least in part at run-time.
Finally, Defendant has also cited prosecution history, but no definitive statements
in this regard are apparent. See Sept. 26, 2011 Appeal Br. [Dkt. # 67-22] at 20–21; see
also Oct. 27, 2011 Office Action [Dkt. # 67-23] at 5–6; Omega Eng’g, 334 F.3d at 1324.
The Court therefore construes “dynamically [selecting / select] one of a plurality
of nodes” to mean “[selecting/select] a node at least in part at run-time based on one
or more variables.”

48 / 71

P. “[associating / associate] a virtual disk image with the selected node”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
this term to mean, “[associating / associate]
a virtual disk file(s) with the selected
node.”

“[storing/store] a pointer or other reference
to a simulated disk drive in persistent storage at the selected node”

[Dkt. # 55] at 7; [Dkt. # 57] at 23; [Dkt. # 67] at 25. The parties submit that these disputed terms appear in Claims 1, 15, and 29 of the ’714 Patent. [Dkt. # 55] at 7.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “[providing/provide] the selected node
with a pointer or other reference to a virtual disk drive.” At the August 4, 2016 hearing,
the parties agreed to the Court’s preliminary construction.
The Court therefore construes “[associating / associate] a virtual disk image
with the selected node” to mean “[providing/provide] the selected node with a pointer or other reference to a virtual disk drive.”

49 / 71

Q. “low utilization”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

Plain and Ordinary Meaning, no construction necessary.
If the Court determines a construction is
required, a PHOSITA would understand
th[is] term[] to mean, “low use.”

[Defendant] alleges that this term is indefinite.

[Dkt. # 55] at 7–8; [Dkt. # 57] at 24; [Dkt. # 67] at 27; [Dkt. # 71] at 9. This disputed
term appears in Claims 6, 20, and 34 of the ’714 Patent. [Dkt. # 55] at 7; [Dkt. # 72-2] at
7.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (not indefinite).”
(1) The Parties’ Positions
Plaintiff argues that the claims and the specification provide context for understanding this relative term. [Dkt. # 57] at 24. Plaintiff also argues that Defendant’s indefiniteness argument should be rejected because it is “made without any extrinsic or testimonial support” and because “[t]he specification shows that low utilization means a node
that is doing less work than other currently-running nodes.” Id. at 25 (citing ’714 Patent
at 9:4–10).
Defendant argues that this disputed term is indefinite because “[i]t is a subjective
determination of the user,” and “[e]ven the metric is unclear.” [Dkt. # 67] at 27.

50 / 71

Plaintiff replies that “[t]he cases that [Defendant] cites deal with the concept of
‘low’ (i.e., low angle or low-efficiency) in the absolute sense, and not in the relative sense
that is used in the ’714 patent.” [Dkt. # 71] at 10.
(2) Analysis
The Supreme Court of the United States has “read [35 U.S.C.] § 112, ¶ 2 to require
that a patent’s claims, viewed in light of the specification and prosecution history, inform
those skilled in the art about the scope of the invention with reasonable certainty.” Nautilus, Inc. v. Biosig Instruments, Inc., 134 S. Ct. 2120, 2129 (2014). “A determination of
claim indefiniteness is a legal conclusion that is drawn from the court’s performance of
its duty as the construer of patent claims.” Datamize, LLC v. Plumtree Software, Inc., 417
F.3d 1342, 1347 (Fed. Cir. 2005) (citations and internal quotation marks omitted), abrogated on other grounds by Nautilus, 134 S. Ct. 2120.
Claim 6 of the ’714 Patent recites (emphasis added):
6. The method of claim 5, further comprising, in response to none of the
nodes being unutilized, selecting one of the nodes utilized by the other distributed application based on one or more of the following:
the retrieved policy;
low utilization of the selected node;
priority of the selected distributed application; or
compatibility of the selected node with the selected distributed application.
The specification demonstrates that the term “low utilization” refers to a relative
degree of utilization:
Returning to decisional step 210, if there were no computing nodes 108
available, then dynamic boot engine 105 selects an optimum utilized node

51 / 71

108 for application 114 at step 250. This selection of optimum node 108
may occur in any appropriate fashion such as, for example, determining the
least utilized node 108, selecting a compatible node 108, or determining
some other “best fit”.
’714 Patent at 9:4–10 (emphasis added).
Defendant’s expert opines that “[t]here are many different ways to measure the
‘utilization’ or ‘use’ of a computer system,” such as “what portion of processor time is
spent executing non-idle threads,” “the amount of available physical or virtual memory,”
“the current network load,” “number of concurrent processes, or the number of current
users logged in.” June 30, 2016 Livny Decl. [Dkt. # 67-3] ¶ 65.
Yet, Defendant’s expert acknowledges that “a particular measure of utilization assessed at one system can be compared with the same measure assessed at another system”
and “can be used to determine which system has a higher level of utilization and which
has a lower level of utilization.” Id. at ¶ 66. Because the claim language and the specification, as set forth above, demonstrate that “low utilization” is used as a relative term rather than an absolute term, Defendant has not demonstrated that the claims at issue fail to
“inform those skilled in the art about the scope of the invention with reasonable certainty.” Nautilus, 134 S. Ct. at 2129; see Interval Licensing LLC v. AOL, Inc., 766 F.3d 1364,
1370 (Fed. Cir. 2014) (“We do not understand the Supreme Court to have implied in Nautilus, and we do not hold today, that terms of degree are inherently indefinite.”).
The Court therefore expressly rejects Defendant’s indefiniteness argument, and the
Court construes “low utilization” to mean “lower utilization.”

52 / 71

R. “the distributed application is operable to execute at a subset of the plurality of
nodes”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

[Plaintiff] disagrees with [Defendant’s] assertion that claims 2 and 30 of the ’714 Patent should be construed under 35 U.S.C.
§112 ¶ 6.
If the Court determines a construction is
required under 35 U.S.C. §112 ¶ 6, [Plaintiff] proposes the following function and
structure:
Function: execute at a subset of the plurality of nodes

This term is a means plus function limitation subject to 35 U.S.C. § 112 ¶ 6 and is
indefinite.
Function: execute at a subset of the plurality of nodes
Structure: Indefinite.
Alternatively, a database management system (DBMS) or financial software. (Col.
4:17–19)

Structure: Dynamic boot engine
[Dkt. # 55] at 8; [Dkt. # 57] at 25. The parties submit that this term appears in Claims 2
and 30 of the ’714 Patent. [Dkt. # 55] at 8.
Plaintiff argued that this disputed term is not a means-plus-function term because
it “merely further limits ‘distributed application’—a term the parties separately construed
and agreed was not subject to 35 U.S.C. § 112 ¶ 6—to having operability to execute at a
subset of the plurality of nodes.” [Dkt. # 57] at 25 (citations omitted). Alternatively,
Plaintiff argued that “[Defendant’s] identification of alternative structures, DBMS or financial software, is overly-narrow because the specification provided those alternative
structures only as non-limiting examples.” Id. at 26 (citing ʼ714 Patent at 4:17–22). In
other words, Plaintiff argued that “[Defendant’s] identification of alternative structures

53 / 71

unduly limits the distributed application to two types and fails to provide structure that
would allow a distributed application to run on a subset of nodes.” [Dkt. # 57] at 27.
Defendant responds that it “no longer seeks construction of this term.” [Dkt. # 67]
at 28; see [Dkt. # 72-2] 3 n.1. The Court therefore does not further address this term.
S. “a manager coupled to a fabric, the manager operable to monitor a currently
running node in the HPC system executing a host and, if a fault occurs at the currently running node, discontinue operation of the currently running node and boot
the host at a free node in the HPC system from the storage”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

[Plaintiff] disagrees with [Defendant’s] assertion that claim 1 of the ’274 Patent
should be construed under 35 U.S.C. §112
¶ 6.
If the Court determines a construction is
required under 35 U.S.C. §112 ¶ 6
Function: monitor the currently running
node and discontinue operation of the currently running node and boot the host at a
free node from storage
Structure: “management node,” which may
include software such as a fault tolerance
and recovery manager

This term is a means plus function limitation subject to 35 U.S.C. § 112 ¶ 6.
Function: monitor the currently running
node and discontinue operation of the currently running node and boot the host at a
free node from storage
Structure: a daemon running on each node
provides a heartbeat to a management node,
and in response to information contained in
the heartbeat, the management node forces
a node to idle or power down, reboots the
node, or prevents communication to the
node by updating one or more routing tables or updating software or other logic at
centralized storage. Booting the host is accomplished by Wake-On LAN, Intelligent
Platform Management Interface (IPMI),
Preboot Execution Environment (PXE), or
Dynamic Host Configuration Protocol
(DHCP).
(Col. 63:14–37, Col. 64:29–45)

54 / 71

[Dkt. # 55] at 8–9; [Dkt. # 57] at 27; [Dkt. # 67] at 28; [Dkt. # 71] at 10; [Dkt. # 72-1] at
3. The parties submit that this disputed term appears in Claim 1 of the ’274 Patent. [Dkt.
# 55] at 8.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “plain meaning (not subject to 35 U.S.C.
§ 112 ¶ 6).”
(1) The Parties’ Positions
Plaintiff argues that Defendant cannot overcome the presumption against meansplus-function treatment, and “[o]nce the court resolves the construction of disputed terms
in this phrase, e.g., ‘node,’ the meaning of the [disputed term] is clear.” [Dkt. # 57] at 27.
Alternatively, Plaintiff argues that the specification discloses a “management node” as
performing the claimed function. Id. at 27–28. Plaintiff also urges that Defendant’s proposal should be rejected because “[Defendant’s] identification contradicts the specification because the management node may run other modules than a daemon to monitor the
system’s health.” Id. at 28 (citation, alteration, and internal quotation marks omitted).
Defendant responds that the presumption against means-plus-function treatment is
overcome because although “the inventor used the word ‘manager,’ . . . the specification
makes clear that the term is simply a name given to any number of software modules in
the patented system.” [Dkt. # 67] at 29.

55 / 71

Plaintiff replies: “[Defendant] does not define the structure based on any specification passages cited in its brief. Instead, [Defendant] relies only on its expert’s declaration.” [Dkt. # 71] at 10.
At the August 4, 2016 hearing, Defendant urged that even if “manager” is deemed
to be a structural term because of its recitation as being “coupled to the fabric,” the
“manager” is nonetheless a generic “black box” that lacks any boundaries.
(2) Analysis
Title 35 U.S.C. § 112, ¶ 6 provides: “An element in a claim for a combination may
be expressed as a means or step for performing a specified function without the recital of
structure, material, or acts in support thereof, and such claim shall be construed to cover
the corresponding structure, material, or acts described in the specification and equivalents thereof.”
“[T]he failure to use the word ‘means’ . . . creates a rebuttable presumption . . .
that § 112, para. 6 does not apply.” Williamson v. Citrix Online LLC, 792 F.3d 1339, 1348
(Fed. Cir. 2015) (citations and internal quotation marks omitted). “When a claim term
lacks the word ‘means,’ the presumption can be overcome and § 112, para. 6 will apply if
the challenger demonstrates that the claim term fails to recite sufficiently definite structure or else recites function without reciting sufficient structure for performing that function.” Id. at 1349 (citations and internal quotation marks omitted).
Williamson, in an en banc portion of the decision, abrogated prior statements that
the absence of the word “means” gives rise to a “strong” presumption against means-

56 / 71

plus-function treatment. Id. (citation omitted). Williamson also abrogated prior statements
that this presumption “is not readily overcome” and that this presumption cannot be overcome “without a showing that the limitation essentially is devoid of anything that can be
construed as structure.” Id. (citations omitted). Instead, Williamson found, “[h]enceforth,
we will apply the presumption as we have done prior to Lighting World . . . .” Id. (citing
Lighting World, Inc. v. Birchwood Lighting, Inc., 382 F.3d 1354, 1358 (Fed. Cir. 2004)).
In a subsequent part of the decision not considered en banc, Williamson affirmed the district court’s finding that the term “distributed learning control module” was a means-plusfunction term that was indefinite because of lack of corresponding structure, and in doing
so Williamson stated that “‘module’ is a well-known nonce word.” Id. at 1350.
Claim 1 of the ’274 Patent, for example, recites (emphasis added):
1. A system for fault tolerance and recovery in a high-performance computing (HPC) system, the system for fault tolerance and recovery comprising:
a fabric coupling a plurality of nodes in an HPC system to each other, each node comprising a switching fabric integrated to a card
and at least two processors integrated to the card;
storage coupled to the fabric and accessible to each of the nodes, the
storage operable to store a plurality of hosts each executable at
any of the nodes; and
a manager coupled to the fabric, the manager operable to monitor a
currently running node in the HPC system executing a host and,
if a fault occurs at the currently running node, discontinue operation of the currently running node and boot the host at a free
node in the HPC system from the storage.
The term “manager” as used here is distinguishable from the word “module” in
Williamson. Id. at 1348 (“What is important is . . . that the term, as the name for structure,

57 / 71

has a reasonably well understood meaning in the art.”) (quoting Greenberg v. Ethicon
Endo-Surgery, Inc., 91 F.3d 1580, 1583 (Fed. Cir. 1996)). Defendant has cited disclosures
in the specification that “[p]hysical manager 505 is any software, logic, firmware, or other module operable to . . .” and “[v]irtual manager 510 is any software, logic, firmware,
or other module operable to . . . .” ’274 Patent at 13:4–5 & 13:15–16. Although the determination of whether 35 U.S.C. § 112, ¶ 6 applies is performed in the context of the
specification, the ultimate inquiry is whether the claim term at issue is structural, not
whether a particular embodiment is characterized in the specification as potentially being
a “module.” See id. Further, the specification also refers to other managers without referring to a “module.” See ’274 Patent at 61:8–12 & 62:33–37; see also Apple Inc. v.
Motorola, Inc., 757 F.3d 1286, 1296 (Fed. Cir. 2014) (“[W]e must construe the claim limitation to decide if it connotes ‘sufficiently definite structure’ to a person of ordinary skill
in the art, which requires us to consider the specification (among other evidence).”), abrogated on other grounds by Williamson, 792 F.3d 1339.
In finding that 35 U.S.C. § 112, ¶ 6 does not apply, the Court applies long-standing
principles articulated prior to the abrogated Lighting World decision. See, e.g., Linear
Tech. Corp. v. Impala Linear Corp., 379 F.3d 1311, 1320 (Fed. Cir. 2004) (“when the
structure-connoting term ‘circuit’ is coupled with a description of the circuit’s operation,
sufficient structural meaning generally will be conveyed to persons of ordinary skill in
the art, and § 112 ¶ 6 presumptively will not apply”; noting “language reciting [the circuits’] respective objectives or operations”); Apex Inc. v. Raritan Computer, Inc., 325

58 / 71

F.3d 1364, 1372 (Fed. Cir. 2003) (“While we do not find it necessary to hold that the term
‘circuit’ by itself always connotes sufficient structure, the term ‘circuit’ with an appropriate identifier such as ‘interface,’ ‘programming’ and ‘logic,’ certainly identifies some
structural meaning to one of ordinary skill in the art.”); Personalized Media Commc’ns,
LLC v. Int’l Trade Comm’n, 161 F.3d 696, 705 (Fed. Cir. 1998) (“Even though the term
‘detector’ does not specifically evoke a particular structure, it does convey to one knowledgeable in the art a variety of structures known as ‘detectors.’ We therefore conclude
that the term ‘detector’ is a sufficiently definite structural term to preclude the application
of § 112, ¶ 6.”); Greenberg, 91 F.3d at 1583 (finding that “detent mechanism” was not a
means-plus-function term because it denotes a type of device with a generally understood
meaning in the mechanical arts).
No further construction is necessary. Indeed, Defendant has not proposed any alternative construction.
The Court therefore construes “a manager coupled to a fabric, the manager operable to monitor a currently running node in the HPC system executing a host and,
if a fault occurs at the currently running node, discontinue operation of the currently running node and boot the host at a free node in the HPC system from the storage” to have its plain meaning.

59 / 71

T. “means for monitoring a currently running node in an HPC system comprising a
plurality of nodes”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

This term is a means plus function limitation subject to 35 U.S.C. § 112 ¶6.

This term is a means plus function limitation subject to 35 U.S.C. § 112 ¶6.

Function: monitor a currently running node
in an HPC system comprising a plurality of
nodes.

Function: monitor a currently running node
in an HPC system comprising a plurality of
nodes.

Structure: “management node,” which may
include software for monitoring currently
running nodes, such as a fault tolerance and
recovery manager.”

Structure: a daemon running on each node
provides a heartbeat to a management node,
and in response to information contained in
the heartbeat, the management node forces
a node to idle or power down, reboots the
node, or prevents communication to the
node by updating one or more routing tables or updating software or other logic at
centralized storage. (Col. 63:25–37)

[Dkt. # 55] at 9–10; [Dkt. # 57] at 28; [Dkt. # 67] at 29–30; [Dkt. # 72-1] at 22. The parties submit that this disputed term appears in Claim 38 of the ’274 Patent. [Dkt. # 55] at
9.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “structure: ‘(1) fault tolerance and recovery manager 538 and a daemon or other software component at node 115 providing a
heartbeat or checkpointing mechanism, and equivalents thereof; or (2) physical manager
505 and agents 132, and equivalents thereof.’”

60 / 71

(1) The Parties’ Positions
Plaintiff submits that the parties agree that this is a means-plus-function term, but
the parties dispute the corresponding structure. [Dkt. # 57] at 28. Plaintiff argues that “[a]
POSITA would understand that the disclosed corresponding structure is the ‘management
node,’ which may include software for monitoring currently running nodes, such as a
fault tolerance and recovery manager.” Id. Plaintiff urges that “[Defendant’s] proposed
identified structure improperly excludes other disclosed embodiments, including the embodiment where the management node may run modules other than a daemon, such as ‘a
fault tolerance and recovery manager 538’ to monitor the HPC system’s health.” Id. at 29
(citing ʼ274 Patent at 62:31–54).
Defendant responds that the corresponding structure must be an algorithm and that
its proposed construction “identifies the patent’s disclosures of the algorithm for performing the claimed function.” [Dkt. # 67] at 30.
Plaintiff replies: “As with the previous term, [Defendant] relies only on its expert’s
declaration for its proposed ¶ 6 structures. [Defendant’s] proposals based only on extrinsic evidence should be rejected.” [Dkt. # 71] at 10 (citation omitted).
(2) Analysis
When it applies, 35 U.S.C. § 112, ¶ 6 limits the scope of the functional term “to
only the structure, materials, or acts described in the specification as corresponding to the
claimed function and equivalents thereof.” Williamson, 792 F.3d at 1347.

61 / 71

Construing a means-plus-function limitation involves multiple steps. “The first
step . . . is a determination of the function of the means-plus-function limitation.” Medtronic, Inc. v. Advanced Cardiovascular Sys., Inc., 248 F.3d 1303, 1311 (Fed. Cir. 2001).
Here, the parties agree that the claimed function is to “monitor a currently running node
in an HPC system comprising a plurality of nodes.”
“[T]he next step is to determine the corresponding structure disclosed in the specification and equivalents thereof.” Id. A “structure disclosed in the specification is ‘corresponding’ structure only if the specification or prosecution history clearly links or associates that structure to the function recited in the claim.” Id. The focus of the “corresponding structure” inquiry is not merely whether a structure is capable of performing the recited function, but rather whether the corresponding structure is “clearly linked or associated with the [recited] function.” Id. The corresponding structure “must include all structure that actually performs the recited function.” Default Proof Credit Card Sys. v. Home
Depot U.S.A., Inc., 412 F.3d 1291, 1298 (Fed. Cir. 2005). However, § 112 does not permit
“incorporation of structure from the written description beyond that necessary to perform
the claimed function.” Micro Chem., Inc. v. Great Plains Chem. Co., 194 F.3d 1250, 1258
(Fed. Cir. 1999).
The specification discloses a “management node 15” that includes a “fault tolerance and recovery manager 538.” ’274 Patent at 62:30–32. “Fault tolerance and recovery
manager 538 includes a hardware, software, or embedded logic component or a combination of two or more such components for detecting faults at nodes 115 in grid 110 and ini-

62 / 71

tiating recovery from such faults.” Id. at 62:32–36 (emphasis added). The specification
then provides additional detail regarding a “heartbeat mechanism”:
In particular embodiments, when a node 115 executes a host, fault tolerance
and recovery manager 538 monitors the host to track health of node 115
and one or more applications running at node 115. In particular embodiments, fault tolerance and recovery manager 538 uses a daemon or other
software component at node 115 providing a heartbeat mechanism to monitor health of node 115. The daemon communicates heartbeat messages to
fault tolerance and recovery manager 538 at regular intervals indicating
whether node 115 is functioning properly. A heartbeat message from the
daemon may provide status information regarding node 115. As an example
and not by way of limitation, a heartbeat message from the daemon may
provide status information regarding node 115 indicating a temperature of
node 115, an average speed of a fan at node 115, and a level of power consumption at node 115. In response to status information indicating a fault at
a node 115, fault tolerance and recovery manager 538 may automatically
and without user input initiate action to recover from the fault, notify an
administrator of HPC server 102 of the fault, or both.
In particular embodiments, fault tolerance and recovery manager 538 considers a node 115 healthy if fault tolerance and recovery manager 538 continues to receive heartbeat messages from node 115 providing status information indicating that node 115 has not exceeded one or more configurable
thresholds. If fault tolerance and recovery manager 538 does not receive a
heartbeat message from node 115 providing status information indicating
that node 115 has not exceeded one or more configurable thresholds, fault
tolerance and recovery manager 538 may automatically and without user
input initiate action to recover from the fault, notify an administrator of
HPC server 102 of the fault, or both.
Id. at 63:22–63:53 (emphasis added); see id. at 63:54–64:12.
This above-quoted passage discloses that structures for monitoring a currently
running node are the “fault tolerance and recovery manager 538” as well as “a daemon or

63 / 71

other software component at node 115 providing a heartbeat mechanism.” See id. at
63:25–28. The specification also discloses a “checkpoint” mechanism, wherein instead of
simply receiving a heartbeat mechanism the fault tolerance and recovery manager 538
actively requests information from nodes. Id. at 64:65–65:16.
Nonetheless, Defendant’s proposal that “the management node forces a node to
idle or power down, reboots the node, or prevents communication to the node by updating one or more routing tables or updating software or other logic at centralized storage”
is rejected as not linked to, or necessary to perform, the claimed function. See, e.g.,
Northrop Grumman Corp. v. Intel Corp., 325 F.3d 1346, 1352 (Fed. Cir. 2003) (“Under
section 112, paragraph 6, structure disclosed in the specification is ‘corresponding’ structure only if the specification or the prosecution history clearly links or associates that
structure to the function recited in the claim. . . . A court may not import into the claim
features that are unnecessary to perform the claimed function.”) (citations and internal
quotation marks omitted).
Finally, the specification also discloses a “[p]hysical manager 505” and “agents
132” that interact so as to “determine the physical health” of nodes:
Physical manager 505 is any software, logic, firmware, or other module operable to determine the physical health of various nodes 115 and effectively
manage nodes 115 based on this determined health. Physical manager may
use this data to efficiently determine and respond to node 115 failures. In
one embodiment, physical manager 505 is communicably coupled to a plurality of agents 132, each residing on one node 115. As described above,
agents 132 gather and communicate at least physical information to man-

64 / 71

ager 505. Physical manager 505 may be further operable to communicate
alerts to a system administrator at client 120 via network 106.
’274 Patent at 13:4–14 (emphasis added). A patent may disclose multiple “alternative
structures for performing the claimed function,” and the Court may identify those alternatives rather than attempt to formulate a single claim interpretation to cover multiple alternatives. Ishida Co., Ltd. v. Taylor, 221 F.3d 1310, 1316 (Fed. Cir. 2000).
The Court therefore finds that “means for monitoring a currently running node
in an HPC system comprising a plurality of nodes” is a means-plus-function term, the
claimed function is to “monitor a currently running node in an HPC system comprising a plurality of nodes,” and the corresponding structure is “(1) fault tolerance and
recovery manager 538 and a daemon or other software component at node 115
providing a heartbeat or checkpointing mechanism, and equivalents thereof; or (2)
physical manager 505 and agents 132, and equivalents thereof.”

65 / 71

U. “means for, if a fault occurs at the currently running node: discontinuing operation of the currently running node; and booting the host at a free node in the HPC
system from the storage”
Plaintiff’s Proposed Construction

Defendant’s Proposed Construction

This term is a means plus function limitation subject to 35 U.S.C. § 112 ¶ 6.

This term is a means plus function limitation subject [to 35] U.S.C. § 112 ¶ 6.

Function: if a fault occurs, discontinuing
operation of the currently running node and
booting the host at a free node from storage.

Function: if a fault occurs, discontinuing
operation of the currently running node and
booting the host at a free node from storage.

Structure: “management node,” which includes software to allow an HPC system to,
if a fault occurs, discontinue operation of
the currently running node and boot a host
at a free node from storage, such as a fault
tolerance and recovery manager

Structure: force a node to idle or power
down, reboot the node, or prevent communication to the node by updating one or
more routing tables or updating software or
other logic at centralized storage. Booting
the host is accomplished by Wake-On
LAN, Intelligent Platform Management Interface (IPMI), Preboot Execution Environment (PXE), or Dynamic Host Configuration Protocol (DHCP). (Col. 64:29–45)

[Dkt. # 55] at 10–11; [Dkt. # 57] at 29; [Dkt. # 72-1] at 23. The parties submit that this
disputed term appears in Claim 38 of the ’274 Patent. [Dkt. # 55] at 10.
Shortly before the start of the August 4, 2016 hearing, the Court provided the parties with the following preliminary construction: “fault tolerance and recovery manager
538; and equivalents thereof.”

66 / 71

(1) The Parties’ Positions
Plaintiff submits that the parties agree that this is a means-plus-function term, but
the parties dispute the corresponding structure. [Dkt. # 57] at 29. Plaintiff argues that
“[Defendant’s] proposed construction improperly omits meaningful structures from its
construction and should be rejected.” Id. at 30.
Defendant responds that the corresponding structure must be an algorithm and that
its proposed construction “identifies the patent’s disclosures of the algorithm for performing the claimed function.” [Dkt. # 67] at 30.
Plaintiff replies: “As with the previous term, [Defendant] relies only on its expert’s
declaration for its proposed ¶ 6 structures. [Defendant’s] proposals based only on extrinsic evidence should be rejected.” [Dkt. # 71] at 10 (citation omitted).
(2) Analysis
Legal principles regarding 35 U.S.C. § 112, ¶ 6 are set forth as to the term “means
for monitoring a currently running node in an HPC system comprising a plurality of
nodes,” above.
As to the present disputed term, the parties agree that the claimed function is: “if a
fault occurs, discontinuing operation of the currently running node and booting the host at
a free node from storage.”
As to the proper corresponding structure, the specification discloses:
In particular embodiments, fault tolerance and recovery manager 538 considers a node 115 healthy if fault tolerance and recovery manager 538 continues to receive heartbeat messages from node 115 providing status infor-

67 / 71

mation indicating that node 115 has not exceeded one or more configurable
thresholds. If fault tolerance and recovery manager 538 does not receive a
heartbeat message from node 115 providing status information indicating
that node 115 has not exceeded one or more configurable thresholds, fault
tolerance and recovery manager 538 may automatically and without user
input initiate action to recover from the fault, notify an administrator of
HPC server 102 of the fault, or both. * * *
In particular embodiments, if a heartbeat message from a daemon at a first
node 115 indicates that a nonrecoverable fault has occurred in hardware executing a host at first node 115 and a configurable script for carrying out
customized recovery with respect to first node 115 specifies a recovery
method that includes restarting the host at a second node 115 in grid 110,
fault tolerance and recovery manager 538 may select a second node 115 for
executing the host and then boot the host at second node 115 for execution
at second node 115. Fault tolerance and recovery manager may then update
one or more routing tables to enable the host to communicate with other
nodes 115 in grid 110 and clients 120 external to HPC server 102. Fault tolerance and recovery manager 538 may indicate in the routing tables or
elsewhere in HPC server 102 that first node 115 is offline and notify an
administrator of HPC server 102 of the fault at first node 115.
In particular embodiments, if fault tolerance and recovery manager 538 has
detected a fault at a first node 115 executing a host and, in response to the
fault, booted a second node 115 and successfully initialized the host at second node 115, fault tolerance and recovery manager 538 may take steps to
discontinue operation of first node 115. As an example and not by way of
limitation, to discontinue operation of first node 115, fault tolerance and recovery manager 538 may update one or more routing tables at HPC server
102 to prevent communication to and from first node 115, update software
or other logic at centralized storage 540 to prevent first node 115 from accessing centralized storage 540, force node 115 to idle or power down, or
reboot first node 115. Discontinuing operation of first node 115 may be
preferable if first node 115 failed not because of a fault in hardware at first
node 115, but because of a network failure or a failure in an OS or other
software at first node 115.

68 / 71

’274 Patent at 63:42–64:45; see id. at 63:14–18 (“To boot the host, fault tolerance and
recovery manager 538 may use one or more of Wake-On LAN, Intelligent Platform Management Interface (IPMI), Preboot Execution Environment (PXE), and Dynamic Host
Configuration Protocol (DHCP).”).
The Court therefore finds that “means for, if a fault occurs at the currently
running node: discontinuing operation of the currently running node; and booting
the host at a free node in the HPC system from the storage” is a means-plus-function
term, the claimed function is “if a fault occurs, discontinuing operation of the currently running node and booting the host at a free node from storage,” and the corresponding structure is “fault tolerance and recovery manager 538; and equivalents
thereof.”
.

V. CONCLUSION
The Court adopts the constructions set forth in this opinion for the disputed terms

of the patents-in-suit. The parties are ordered to not refer to each other’s claim construction positions in the presence of the jury. Likewise, in the presence of the jury, the parties
are ordered to refrain from mentioning any portion of this opinion, other than the actual
definitions adopted by the Court. The Court’s reasoning in this order binds the testimony
of any witnesses, but any reference to the claim construction proceedings is limited to informing the jury of the definitions adopted by the Court.
SIGNED this 3rd day of January, 2012.

SIGNED this 9th day of June, 2017.

____________________________________
ROY S. PAYNE
UNITED STATES MAGISTRATE JUDGE

69 / 71

APPENDIX A
Parties’ Agreement

Term
“processor(s)”
(’909 Pat., Cls. 1–7, 15–21, 23, 25–30, 32)
(’833 Pat., Cls. 1–6, 8–13)
“communicably coupled,” “communicably
couple,” and “communicably couples”
(’909 Pat., Cls. 1, 4–6, 15, 20, 23, 27–29)
(’833 Pat., Cls. 1, 8, 11, 12)
“couples”/“coupling”

The Parties understand this term to mean
“device that executes instructions.”

The Parties understand this term to mean
“connected / connect / connects, to permit
communications.”

The Parties understand this term to mean
“connects / connecting.”

(’909 Pat., Cls. 2, 3, 15–19, 25–26)
(’833 Pat., Cls. 2, 3, 4, 9, 10)

The Parties understand this term to mean
“adapter that provides channel-based connections.”

“Host Channel Adapter (HCA)”
(’833 Pat., Cls. 2 and 9)

The Parties understand this term to mean
“switch comprises twenty-four ports and
enables a toroidal topology comprising four “switch with at least twenty-four ports and
forms a four dimensional topology where
dimensions”
each node connects to exactly eight neigh(’833 Pat., Cls. 7 and 14)
bors.”
“peripheral component interconnect (PCI)
bridge”
(’833 Pat., Cls. 3 and 10)
“daemon”
(’274 Pat., Cls. 2, 4, 5, 17, 19, 20, 28, 30,
31)
“checkpoint”
(’274 Pat., Cls. 6, 21, 32)

The Parties understand this term to mean
“circuit that interfaces a PCI or PCI Express bus/link to another bus or link.”
The Parties understand this term to mean
“background program.”
The Parties understand this term to mean
“snapshot.”

70 / 71

“routing table”
(’274 Pat., Cls. 7, 22, 33)
“timing of the selection of the node”
(’714 Pat., Cls. 9, 23, 37)
“three dimensional torus”
(’274 Pat., Cls. 10, 25, 36)

The Parties understand this term to mean
“table of network addresses and/or ports to
route data.”
The Parties understand this term to mean
“calendar date or time of the selection of
the node.”
The Parties understand this term to mean
“three dimensional topology where each
node connects to exactly six neighbors.”

[Dkt. # 55] at 2–3.

71 / 71