Oracle America, Inc. v. Google Inc.

Filing 493

MOTION in Limine NO. 2 filed by Google Inc.. Responses due by 10/21/2011. Replies due by 10/28/2011. (Attachments: #1 Oracle Opposition)(Kamber, Matthias) (Filed on 10/7/2011)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

MORRISON & FOERSTER LLP
MICHAEL A. JACOBS (Bar No. 111664)
[e-mail]
MARC DAVID PETERS (Bar No. 211725)
[e-mail]
DANIEL P. MUINO (Bar No. 209624)
[e-mail]
755 Page Mill Road, Palo Alto, CA 94304-1018
Telephone: [phone] / Facsimile: [phone]
BOIES, SCHILLER & FLEXNER LLP
DAVID BOIES (Admitted Pro Hac Vice)
[e-mail]
333 Main Street, Armonk, NY 10504
Telephone: [phone] / Facsimile: [phone]
STEVEN C. HOLTZMAN (Bar No. 144177)
[e-mail]
1999 Harrison St., Suite 900, Oakland, CA 94612
Telephone: [phone] / Facsimile: [phone]
ORACLE CORPORATION
DORIAN DALEY (Bar No. 129049)
[e-mail]
DEBORAH K. MILLER (Bar No. 95527)
[e-mail]
MATTHEW M. SARBORARIA (Bar No. 211600)
[e-mail]
500 Oracle Parkway, Redwood City, CA 94065
Telephone: [phone] / Facsimile: [phone]
Attorneys for Plaintiff
ORACLE AMERICA, INC.

18

UNITED STATES DISTRICT COURT

19

NORTHERN DISTRICT OF CALIFORNIA

20

SAN FRANCISCO DIVISION

21

ORACLE AMERICA, INC.

22

Plaintiff,

23

v.

24

GOOGLE INC.

25
26
27

Defendant.

Case No. CV 10-03561 WHA
ORACLE AMERICA, INC.’S
OPPOSITION TO GOOGLE’S
MOTION IN LIMINE NO. 2 TO
EXCLUDE ALLEGED
PERFORMANCE BENEFITS OF
ANDROID WITH THE ACCUSED
FUNCTIONALITY
Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup

28
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential

1
2

INTRODUCTION
It is indisputable that performance is crucial to the commercial success and adoption of

3

Android. Google devotes an entire section of its Android website discussing performance

4

optimizations (see, e.g., Agrawal Decl. Ex. 2-1); it publicly touts Android performance highlights

5

in another section (see, e.g., id. Ex. 2-2); and it understands that performance is of utmost

6

importance to the success of Android (see, e.g., id. Ex. 2-3, GOOGLE-04-00055098 (“if the

7

device is not fast and stable we FAIL”); Ex. 2-4 (Morrill Dep. 168:22-169:23 (Android engineer

8

testified that “In other words, the device’s poor performance would reflect -- in the user’s

9

eyes”)).)

10

Based on Google’s own evidence, Prof. Mitchell described how Google’s inclusion of the

11

claimed inventions of five patents-in-suit in Android was performance-motivated, citing to a 2008

12

Google I/O presentation entitled “Dalvik Virtual Machine Internals” by Dan Bornstein. (Agrawal

13

Decl. Ex. 2-5 (Mitchell Patent Infringement Report) ¶ 83.) To evaluate the performance impact

14

of the claimed inventions of the patents-in-suit on Android, Prof. Mitchell asked that Android’s

15

performance be tested. (Id. ¶ 6.) Prof. Mitchell worked with Java engineers Bob Vandette, Noel

16

Poore, Erez Landau, Peter Kessler, and Seeon Birger to figure out how to conduct performance

17

tests to measure the benefit the patented technology confers on Android. (Id. ¶¶ 6, 39-67);

18

Agrawal Decl. Ex. 2-6 (Mitchell Patent Dep. 80:6-9, 81:8-15).)

19

Bob Vandette, Noel Poore, and Erez Landau are qualified engineers with years of

20

experience conducting performance analyses of Java technologies. (Agrawal Decl. Ex. 2-7

21

(Summary and Report of Robert G. Vandette (“Vandette Report”) ¶ 11); Ex. 2-8 (Summary and

22

Report of Noel Poore (“Poore Report”) ¶ 11); Ex. 2-9 (Summary and Report of Erez Landau

23

(“Landau Report”) ¶¶ 9, 11).) Prof. Mitchell, who had determined which portions of the Android

24

source code were responsible for Google’s infringement and described his analysis in a 387-page

25

report and in claim charts spanning 464 pages, advised the engineers about what source code

26

modifications would be needed to disable Android’s infringing functionality. (Agrawal Decl.

27

Ex. 2-5 (Mitchell Patent Infringement Report ¶¶ 40, 52, 60).) The engineers then modified

28

Android in the manner directed by Prof. Mitchell and compared the performance of the
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

1

1

unmodified, infringing Android to the modified Android, occasionally enlisting help from

2

colleagues with whom they normally work in the ordinary course of their performance analysis

3

work. (See, e.g., id. Ex. 2-10 (Landau Dep. 108:13-109:10).) The engineers conducted the

4

Android performance testing just like they conduct Java performance testing:

5

Q. Speaking more generally about the tests that you and your team ran, would
you say that the tests that are described in your report and you’ve testified about
today, are they representative of the kind of performance testing that you and your
team members do in your everyday work?

6
7

A. Yes. This is a typical scenario of work that we are doing.

8

Q. Okay. And when you do your work, do you normally rely on your colleagues
like Seeon to help you do your work?

9
10

A. I assist a lot of and -- I am being assisted a lot with hands-on work with other
of the -- I’m supervising quite multiple projects.

11

Q. Is there anything about the projects that you did in your analysis of Android
that you would say is not typical of the kinds of testing you do in the Java
environment?

12
13

A. No. We use normal methods that -- normal procedure.

14
15

(Id.)
As Mr. Landau explained in his deposition, Seeon Birger assisted him. In this instance,

16
17
18
19
20
21
22

the work that Mr. Birger performed was that of a technician under Mr. Landau’s direction;
Mr. Birger made no decision as to what Android functionality should be disabled or how it should
be disabled, and Mr. Birger did not offer any evaluation of any test results. For a completely
different task, Mr. Birger also assisted in making Android builds that could be loaded onto Nexus
One Android devices for testing purposes by other experts; but here again, Mr. Birger performed
1
routine tasks rather expert analysis. (Birger Summary of Investigation for Damages Expert

(“Birger Summary”) ¶ 11.) Mr. Birger described his later work in a summary that was served
23
with Oracle’s damages report; the work described in Mr. Birger’s summary was not the
24
foundation of the Vandette, Poore, and Landau reports. (See, e.g., Poore Report ¶ 25; Agrawal
25
Decl. Ex. 2-12 (Poore Dep. 24:6-9).) Google is mistaken when it argues otherwise. (Google’s
26
1

27
28

There are no contradictions about who did what, contrary to Google’s argument. (Google’s
MIL #2 at 3.) What the engineers testified they did, they did. And there was no cause to depose
Mr. Birger because he is not a testifying expert and fact discovery had closed.
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

2

1

MIL #2 at 3.) Prof. Kemerer took the unmodified and modified Android builds and had them

2

installed on Nexus One smartphones, and then ran the Linpack performance benchmark on each

3

one. (See Kemerer Decl ¶ 8.) Linpack is a benchmark that Google itself uses to boast about

4

Android performance gains. (Agrawal Decl. Ex. 2-2.) Oracle’s damages expert Prof. Cockburn

5

used these test results in his damages analyses.

6

The performance tests that Oracle’s experts conducted followed this methodology:

7

(1) download and build Android source code following the instructions that
Google makes available to the public via
http://source.android.com/source/index.html (Landau Report ¶ 20; Vandette
Report ¶¶ 13, 22; Poore Report ¶¶ 13, 55);

8
9
10
11
12
13
14
15
16
17

(2) keep an unmodified version of a particular Android release (Froyo 2.2) to
serve as a basis for comparison (Landau Report ¶ 25; Poore Report ¶¶ 13, 55;
Vandette Report ¶¶ 13, 56);
(3) disable the accused functionality in a copy of the same version of Android
(Landau Report ¶ 26; Poore Report ¶¶ 15-17, 57; Vandette Report ¶ 13);
(4) select a hardware test platform that supported Android (such as Google’s
Android emulator, test devices like the Beagleboard or Tegra board, or
Android phone like the Nexus One) (Landau Report ¶¶ 30-31; Poore Report ¶
39; Vandette Report ¶ 19);
(5) run and repeat tests on a device running the unmodified version of Android
(Landau Report ¶¶ 30-31; Poore Report ¶¶ 15-17, 57; Vandette Report ¶ 47);

18

(6) run and repeat the same tests on the same or duplicative device running the
modified version of Android. (Landau Report ¶¶ 30-31; Poore Report ¶¶ 1517, 57; Vandette Report ¶ 47); and

19

(7) report on the results.

20

Google does not dispute that this was the methodology the performance experts employed.

21

(Google’s MIL #2 at 1-2.) Each performance expert explained his work in a report or statement

22

submitted to Google, attached the applications or benchmarks that were used as tests, attached the

23

source code modifications, and attached the results, favorable or not.

24

What the performance test results showed was astounding:

25

Android’s use of the ’104 patent’s claimed invention resulted in as much as 13X
execution speed improvement;
Android’s use of the ’205 patent’s claimed invention resulted in as much as 3.3X
execution speed improvement;
Android’s use of the ’702 patent’s claimed invention resulted in as much as 3.3X .dex file
size improvement; and

26
27
28

ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

3

1
2
3
4
5
6
7

The ’720 patent’s claimed invention resulted in as much as 40% memory savings and also
a 0.10 second savings per application launch time.
(Agrawal Decl. Ex. 2-5 (Mitchell Patent Infringement Report ¶¶ 41, 48, 53, 61).) The test results
for the ’520 patent depended on the nature of the application being tested. (Poore Report ¶¶ 7378.) The ’447 and ’476 patents were not tested.
The benchmarks the performance experts used are the same as those that Google itself
uses to evaluate Android’s performance. Google’s webpage (id. Ex. 2-2) highlights the “Dalvik
Performance Boost” achieved by using JIT technology that infringes the ’205 patent:

8
9
10
11
12
13
14
15
16
17
18

Notably, three of the four benchmarks Google identifies on its own webpage – CaffeineMark,

19

Linpack, and SciMark – were used by Oracle’s performance experts to assess the performance

20

impact of the claimed inventions of the patents-in-suit that Google now attempts to exclude. (See,

21

e.g., Vandette Report ¶ 16 (CaffeineMark and SciMark); Kemerer Decl. ¶ 8 (Linpack).)

22

Moreover, the performance tests, results, and interpretation were consistent with ones Google’s

23

own engineers reported (see, e.g., id. Ex. 2-13 (reporting on “2x to 5x” performance

24

improvements due to JIT technology that is accused of infringing the ’205 patent).) Other

25

examples include:

26
27

Mr. Vandette conducted performance tests on a Tegra board (Vandette Report ¶ 19);
Google itself has used Tegra boards to analyze Android performance (see, e.g., Agrawal
Decl. Ex. 2-14, GOOGLE-02-00465974 (Android execution speed performance
discussion concerning “producing the numbers on our Tegra board”));

28
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

4

1

distributes because of the performance consequences, and they show Google’s motivation for

2

including Oracle’s patented technologies in Android. They are relevant to damages as well, not

3

only because they show the value of the patented inventions, but also because they inform Prof.

4

Cockburn’s econometric and conjoint analyses. Oracle’s performance tests satisfy Rule 403 as

5

well—no unfair prejudice arises from proof of exactly how valuable infringement is to Google.

6

Google pursues this motion in limine precisely because the performance testing evidence

7

is so powerful. Google offers a host of criticisms of Oracle’s performance testing. None

8

establish that the testimony is unreliable. Google’s criticisms go only to weight, not

9

admissibility. Rather than exclusion, cross-examination of Oracle’s witnesses is appropriate

10

means for Google to address its complaints.

11

I.

12

The question for the Court is whether Oracle’s performance testing was reliable enough to

13
14
15
16
17
18
19
20
21
22
23

GOOGLE FAILS TO MEET THE THRESHOLD FOR EXCLUSION OF
EXPERT EVIDENCE AND TESTIMONY

be admissible. Prof. Mitchell and the performance experts are eminently qualified to have
performed the work, and Google does not argue otherwise. Thus, the sole issue for this motion is
whether the reasoning and methods underlying the performance tests are reliable, and whether
they have been properly applied to the facts, pursuant to Rule 702. They have. Google’s
attorney-argument-based criticisms do not establish that Oracle’s methodology was fatally
flawed. Gutierrez v. Wells Fargo & Co., No. C 07-05923 WHA, 2010 WL 1233810, at *11
(N.D. Cal. Mar. 26, 2010) (“[T]he rules of evidence do not demand perfection. Rather, a court
need only determine whether the reasoning and methods underlying the expert testimony are
reliable, and whether they have been properly applied to the facts.”). Google’s motion should be
denied.
A.

24

Oracle’s Performance Testing Satisfies Each Daubert Factor.
1.

25

The Performance Tests and Results Are Based Upon Sufficient
Facts and Data

26

Oracle’s performance experts started by downloading and building Android source code

27

following Google’s own instructions. (See Landau Report ¶¶ 22, 23; Vandette Report ¶¶ 13, 22;

28

Poore Report ¶¶ 13, 55.) They used hardware platforms that were either supported by Android or
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

6

1

had Android distributions. (See, e.g., Vandette Report ¶¶ 20-21.) The benchmarks used to test

2

Android performance were “industry accepted benchmarks” (see, e.g., id. ¶¶ 15-16) that Google

3

itself adopted. In some cases, “[t]he benchmarks used to test the performance of the workspaces

4

were all Google standard applications available in the repository.” (Landau Report ¶ 29.) The

5

experiments comprising the performance benchmark testing were repeated several times, (see,

6

e.g., Poore Report ¶¶ 25, 27), and were consistent with Google’s own touted benchmark results.

7

Thus, the performance tests and results are based upon sufficient facts and data.

8
9

Google challenges the performance reports on the grounds that Prof. Mitchell did not
actually conduct the tests himself; likewise, Google takes issue with Vandette and Landau

10

receiving assistance in conducting performance tests. But “an expert need not have obtained the

11

basis for his opinion from personal perception.” Monsanto Co. v. David, 516 F.3d 1009, 1015

12

(Fed. Cir. 2008). An expert may instruct others on how to conduct the testing.

13

2.

14
15

The Performance Tests and Results are the Product of Reliable
Principles and Methods

Oracle’s performance experts used reliable principles and methods in conducting the

16

performance testing. In Daubert, the Court held that “[w]idespread acceptance can be an

17

important factor in ruling particular evidence admissible.” Daubert v. Merrell Dow Pharms.,

18

Inc., 509 U.S. 579, 594 (1993). In this case, the performance tests are the product of the “widely

19

accepted industry practice to use performance benchmark analysis for the purpose of evaluating

20

the relative value of one product over another.” (Vandette Report ¶ 12.) In his report,

21

Mr. Vandette discusses that performance benchmark analysis is used in the computer science

22

industry and provides examples of some accepted industry benchmarks. (Id. ¶¶ 12-15.) This

23

widespread acceptance of performance benchmark testing weighs in favor of admissibility,

24

demonstrating that performance benchmark analysis is an accepted and reliable methodology.

25

See Daubert, 509 U.S. at 594.

26

Oracle’s performance experts adopted performance measurement techniques and

27

interpretations that Google itself has used to evaluate Android’s performance, supra. Google’s

28

argument that Oracle’s experts offered “no recognized methodology guiding their modifications
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

7

1

of the Android source code” is nonsense. (Google’s MIL #2 at 4.) Disabling a software feature

2

by editing the source code and comparing its performance to the original code using standard

3

benchmarks is not any different from adding a software feature by editing the source code and

4

comparing its performance to the original code using standard benchmarks, which is what

5

Oracle’s performance experts do every day. (See Vandette Report ¶ 12 (Benchmarks used in

6

testing “are written to allow quick and repeatable measurements of a function or set of functions

7

that a platform or product might implement.”).)

8

Oracle’s performance experts selected benchmarks “that did not require any Java

9

graphical APIs and could run on the APIs that are implemented in Android.” (Id. ¶ 16.) Oracle’s

10

experts conducted the performance analysis in such way as to “also allow Google to validate

11

these benchmark tests if they choose to.” (Id. ¶¶ 20, 21.) Thus Oracle’s performance tests are

12

based on reliable methodology. (Id. ¶¶ 5-20.)

13

At trial, Prof. Mitchell and Oracle’s performance experts will testify about why the

14

modifications made were reasonable and reliable to measure the impact of the claimed inventions

15

of the patents-in-suit. Here are some reasons why the Android test modifications were reasonable

16

and reliable, in specific response to Google’s criticisms (Google’s MIL #2 at 4):

17
18
19
20
21
22
23
24
25
26
27
28

’104 patent test: Mr. Vandette disabled Android’s JIT feature in both the unmodified and
modified test code to test the impact of the claimed invention of the ’104 patent.
(Agrawal Decl. Ex. 2-11 (Vandette Dep. 84:21-85:8); Vandette Report ¶¶ 36-39).)
Because Android’s JIT feature depends on the ’104 patented technology, Mr. Vandette
had no choice but to turn off the JIT feature. (Agrawal Decl. Ex. 2-11 (Vandette Dep
84:21-85:8).) In doing so, Mr. Vandette compared apples to apples and reported on the
differences in execution speed.
’205 patent test: Oracle accuses Android’s JIT feature of infringing the ’205 patent. As
Mr. Vandette testified, Android’s JIT feature necessarily relies on Android’s JIT compiler;
and so turning off the JIT compiler was sufficient to turn off Android’s JIT without
affecting other Android features. (Agrawal Decl. Ex. 2-11 (Vandette Dep. 100:9-17).)
’702 patent test: Mr. Poore used three different methodologies to estimate the effect of
the use of the ’702 invention on the size of Android’s Dalvik executable file (dex file).
(Poore Report ¶¶ 15-17.) Mr. Poore’s analysis consistently showed that a dex file would
be larger without benefit of the claimed invention. Using multiple methodologies to
confirm results is generally considered good science. In one of three methodologies,
Mr. Poore explained how it created larger entries, but also explained how he subtracted
out the additional size to get an accurate result. (Id. ¶ 20.)
’520 patent test: Apparently less significant to Google, Google offers no explicit criticism
of Mr. Poore’s ’520 performance tests. To be clear, Mr. Poore turned off the’ 520 patent
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

8

1

by setting an environment variable NO_ARRAY_OPT, which precisely maps to turning
off the patented feature of the ’520 patent. (Id. ¶¶ 59-61.)

2
3
4

’720 patent test: The ’720 patent necessarily relies on a feature called copy-on-write; and
so turning off Android’s copy-on-write feature was sufficient to turn off Android’s use of
the ’720 patent to measure its performance impact. (Landau Report ¶ 27; Agrawal Decl.
Ex. 2-10 (Landau Dep. 46:2-47:8).)

5

Oracle’s performance experts did the exact opposite of what Google accuses them of – they

6

applied careful reasoning to determine how to modify Android source code to fairly reflect

7

disablement of the claimed inventions in order to measure their performance impact. Notably,

8

Google’s damages expert Gregory Leonard contends that disabling the inventions is an alternative

9

to infringement. Lastly, Prof. Mitchell answered Google’s criticisms in his Reply Patent

10

Infringement Report, and will testify to the same.

11

3.

12

Oracle’s Performance Experts Applied the Principles and
Methods Reliably to the Facts of the Case

The principles and methods of performance benchmark analysis were applied reliably to
13
the facts of this case. The engineers repeated their experiments multiple times and performed the
14
experiments on multiple devices. (See Poore Report ¶ 25; Landau Report ¶ 31.) Each engineer
15
provided detailed records of what he did and how he did it. (See, e.g., Vandette Report ¶¶ 26-39;
16
Poore Report ¶¶ 15-40; Landau Report ¶¶ 25-36.) Prof. Mitchell’s report shows that “Google’s
17
own performance benchmarking shows consistent results as the performance benchmarking
18
conducted at [his] request and under [his] direction.” (Agrawal Decl. Ex. 2-5 (Mitchell Patent
19
Infringement Report ¶ 92).) Oracle’s experts have presented their results to Google—results that
20
are reproducible and follow standard practices in the field. (See, e.g., Poore Report ¶¶ 15-40;
21
Landau Report ¶¶ 25-36; Vandette Report ¶¶ 26-39.)
22
It bears repeating: Oracle’s experts conducted the performance analysis in such way as to
23
24
25
26
27
28

“also allow Google to validate these benchmark tests if they choose to.” (Vandette Report ¶ 20,
21.) Google could have repeated the same experiments to determine whether the results were
reproducible and accurate. Google chose not to do so.
Oracle’s experts were not required to run the alternative tests Google’s counsel argues
after-the-fact should have been run (but did not run itself) to rule out every conceivable
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

9

1

performance factor. Hemmings v. Tidyman’s Inc., 285 F.3d 1174, 1188 (9th Cir. 2002) (holding

2

expert testimony admissible even though expert “did not ‘eliminate all of the possible legitimate

3

nondiscriminatory factors’”). Even if the testing did not cover all possible factors, this is a

4

question going to the weight of the evidence, not its admissibility. See id. (“In other words, in

5

most cases, objections to the inadequacies of a study are more appropriately considered an

6

objection going to the weight of the evidence rather than its admissibility.”). Google is

7

attempting to hold Oracle to a “near-impossible standard” of eliminating each and every

8

alternative possibility. See id. This is legally incorrect.

9

Google claims that the performance reports “are so fraught with errors, assumptions and

10

an overall lack of detail that, had the reports been submitted for peer review to any reputable

11

technical journal or academic institution, they would be immediately rejected.” (Google’s MIL

12

#2 at 2.) Google offers no evidence to support this bold assertion. Nor does Google have a solid

13

legal basis for it. The one case it cites, Ellis v. Costco Wholesale Corp., No. 07-15838, 2011 U.S.

14

App. LEXIS 19060 (9th Cir. Sept. 16, 2011), does not address the issue of peer review as a

15

Daubert factor, id. at *25-26, and does not even concern the admissibility of expert testimony.

16

See id. at *20-21. As to the issue of peer review, although it is a factor under Daubert, it is not

17

dispositive. Daubert, 509 U.S. at 594 (“The fact of publication (or lack thereof) in a peer

18

reviewed journal thus will be a relevant, though not dispositive, consideration in assessing the

19

scientific validity of a particular technique or methodology on which an opinion is premised.”).

20

Google’s failure to submit any evidence (in the form of a declaration from a knowledgeable

21

person) to suggest that Oracle’s performance tests would have been accepted or rejected by peer

22

review, or even that software performance testing of this kind is normally submitted to peer-

23

reviewed journals, leaves Google’s argument wholly without foundation.

24

At trial, Prof. Mitchell and Oracle’s performance experts will testify about why their

25

conclusions are supported by the performance tests. Oracle will respond briefly here to Google’s

26

criticisms, in order (Google’s MIL #2 at 4-5):

27
28

Mr. Vandette did exactly what Google did to measure the performance speedup due to
new technologies in Android (see, e.g., Agrawal Decl. Ex. 2-2). Oracle’s performance
experts will testify that benchmarks are widely used as proxies to measure the
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

10

1
2
3
4
5
6

performance of applications. (See Vandette Report ¶ 12.) Thus, it is inaccurate to say that
Mr. Vandette did not test the performance of any actual Android applications. Moreover,
Mr. Vandette used hardware that Google used internally to test Android. The nVidia
Tegra-2 board and TI Beagleboard are made by the companies that make the CPUs for
Android phones. See also Liquid Dynamics Corp. v. Vaughan Co., 449 F.3d 1209, 1221
(Fed. Cir. 2006) (rejecting a challenge to the admissibility of expert opinion on the
grounds that errors in modeling parameters rendered the test results inadmissible); see
also id. at 1220. In Liquid Dynamics, the court rejected defendant’s argument that
computer simulation models were inadmissible because the modeling parameters “did not
match the real-world parameters,” finding instead that the modeling “based on reliable
scientific methodology and subject to cross examination and the proffering of further
scientific analysis by [defendant].” Id. at 1220, 1221.

7
8

Mr. Poore carefully studied dexdump files from an actual Android device to measure the
impact of the ’702 and ’520 patents. (Poore Report ¶ 17.) Thus, it is inaccurate to say
that Mr. Poore did not test the performance of any actual Android applications.

9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

Mr. Landau modified Android source code in order to test Android’s use of the ’720
patent. Google’s criticism of the Nexus One as a test device is belied by Google’s own
August 2010 announcement, discussed above, that the Nexus One was a “good choice” for
experimenting. (Agrawal Decl. Ex. 2-17.)
Google’s position here on “reasonable” non-infringing substitutes is directly contradicted
by Google’s damages expert Gregory Leonard. He opined that a non-infringing
alternative was simply to not include the accused functionality, thereby validating
Oracle’s testing methodology.
It’s not clear what “overall system performance” testing Google refers to—the engineers
used the same benchmarks that Google uses internally and externally. The engineers’
comparisons were apples-to-apples, and therefore methodologically sound. Moreover, it
is not more accurate to recognize that the Dalvik VM is used only 1/3 of the time by
Android, Mr. Vandette testified. (Id. Ex. 2-11 (Vandette Dep. 43:6-44:10 (in reference to
Vandette Dep. Ex. 462)).) Notably, Google doesn’t explain why the Dalvik 1/3 issue
would have any effect on the reliability of the testing. Indeed, Google touts the
performance advantage Dalvik obtains from the infringing JIT on its website, without
accounting for any “1/3 of the time” use. (Agrawal Decl. Ex. 2-2.)
The engineers followed Google’s instructions to download the Android source code from
the public repository maintained by Google. If the code that Google distributes to its
handset maker licensees is different, that is a fact that Google has concealed in discovery.
Android’s Andy Rubin caused a stir in 2010 when he responded to Apple’s Steve Jobs’s
criticism of alleged Android openness by announcing: “the definition of open: ‘mkdir
android ; cd android ; repo init -u git://android.git.kernel.org/platform/manifest.git ; repo
sync ; make’”. The download location Mr. Rubin identified is the same one identified in
Google’s Android download instructions, and the one the engineers used.
The engineers stated in their reports and will testify at trial that they followed Google’s
specific instructions. Mr. Landau was mistaken at his deposition: JDK5 was used for
building the code used in his tests, as he personally confirmed when he returned to his
office in Israel afterwards. His deposition errata will include this correction.
Disputes about these matters are not properly resolved on a motion in limine. Instead, they are
matters for the jury to evaluate.

27
28
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

11

1

B.

Google’s Motion in Limine is Not an Appropriate Vehicle to Contest
the Performance Evidence

2
3

Google’s criticisms of Oracle’s performance evidence go to its weight, not admissibility.

4

As the Liquid Dynamics court held, “[t]he identification of such flaws in generally reliable

5

scientific evidence is precisely the role of cross-examination.” 449 F.3d at 1221 (quoting Quiet

6

Tech. DC-8, Inc. v. Hurel-Dubois UK Ltd., 326 F.3d 1333, 1345 (11th Cir. 2003)). Google’s

7

challenges to the performance reports are inappropriate for a motion in limine. Instead, the

8

alleged errors and deficiencies should be met by “competing expert testimony and active cross-

9

examination.” Id. at 1221; see also Yu-Santos v. Ford Motor Co., No. 1:06-CV-01773-AWI-

10

DLB, 2009 U.S. Dist. LEXIS 41001, at *26 (E.D. Cal. May 13, 2009) (“In most cases, objections

11

to the inadequacies of a study are more appropriately considered objections going to the weight of

12

the evidence rather than their admissibility. If Defendants wish to explore issues that [the expert]

13

did not test for, such as why he did not test the actual seat belt components, these issues can be

14

covered by cross-examination.”) (internal citations omitted); Hemmings, 285 F.3d at 1188

15

(“[O]bjections to the inadequacies of a study are more appropriately considered an objection

16

going to the weight of the evidence rather than its admissibility.”). Google’s attorney-argument-

17

criticisms usurp the role of the jury in determining the weight and persuasiveness of expert

18

testimony. Yu-Santos, 2009 U.S. Dist. LEXIS 41001, at *25 (“Analysis of the credibility and

19

weight of the experts’ conclusions should be reserved for the trier of fact.”). Evidence cannot be

20

excluded simply because Google’s counsel asserts it is flawed or lacking in some regard. The

21

performance reports satisfy the evidentiary requirements; any alleged flaw or defect in the reports

22

must be evaluated by the jury. See id.

23

II.

24
25

THE PERFORMANCE EVIDENCE HAS SIGNIFICANT PROBATIVE
VALUE ON THE QUESTION OF LIABILITY

Google asserts, without any evidence or even explanatory argument, that “evidence

26

regarding the purported performance impact of the asserted patents is potentially probative of

27

only damages theories.” (Google MIL #2 at 5-6.) Google is wrong. Oracle’s performance tests

28

have significant probative value on the question of liability:
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

12

1

7

The performance tests are probative of liability and willful infringement because they
show Google’s motivation for including Oracle’s patented technologies in Android.
The performance tests are probative of direct infringement because they show that the
accused code does run on Android devices. Indeed, these tests demonstrate that the
accused functionality is executed on Android devices and can be disabled. (See, e.g.,
Vandette Report ¶¶ 27-28.)
The performance tests are probative of direct and indirect infringement because they show
that the use of the infringing functionality occurs as a result of following the instructions
Google provides. (See id. ¶ 22.)
The performance tests are probative of indirect infringement because they show that thirdparty device manufacturers are unlikely to make modifications to the Android source code
that Google provides to them because of the performance consequences.

8

Google also argues that the performance analysis would unfairly prejudice Google.

9

Though this evidence may be damaging to Google’s case, the resulting prejudicial effect is hardly

2
3
4
5
6

10

unfair. “That evidence may decimate an opponent’s case is no ground for its exclusion under

11

403. The rule excludes only evidence where the prejudice is ‘unfair’—that is, based on

12

something other than its persuasive weight.” United States v. Cruz-Garcia, 344 F.3d 951, 956

13

(9th Cir. 2003). There is nothing “inflated” about the importance of Oracle’s patents to the

14

success of Android as demonstrated by the performance testing. If anything else, it is Google’s

15

failure to perform testing of its own that may cause the jury to credit Oracle’s performance

16

testing. But that is not unfair—and any alleged “inflation” of that importance can be addressed

17

through Google’s cross examination.

18

III.

19

THE PERFORMANCE EVIDENCE HAS SIGNIFICANT PROBATIVE
VALUE ON THE QUESTION OF DAMAGES

20

The performance testing evidence has significant probative value on the question of

21

damages. The evidence demonstrates a marked impact on Android’s performance when the

22

accused functionality is disabled. (Agrawal Decl. Ex. 2-5 (Mitchell Patent Infringement Report

23

¶¶ 41, 48, 53, 61).) Google admits internally that Android’s performance is crucial to its success.

24

(See, e.g., Agrawal Decl.

25

169:23.)

26
27

); Ex. 2-4 (Morrill Dep. 168:22-

The benchmark tests are particularly relevant to the damages apportionment analysis,
which the Court required Prof. Cockburn to estimate in its Daubert Order. Prof. Cockburn

28
ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

13

1

considers the performance of the modified phones, along with the evidence of the importance of

2

those performance attributes from the conjoint analysis, to calculate the reduced market share that

3

a phone without the patented functionality would command. He used the information from the

4

Linpack benchmarking in the econometrics analysis to calculate customers’ willingness to pay for

5

the patented features. Both of these analyses will help the jury understand the value of the

6

patents-in-suit and enable them to determine damages specific to the infringed patents.

7

Google claims that the performance evidence will “be used in prejudicial ways.” (Google

8

MIL #2 at 6.) But there is no danger of unfair prejudice. As discussed above, unfair prejudice

9

requires more than just the fact that evidence is unfavorable; it requires that prejudice be based on

10

something other than the persuasive weight of the evidence. Cruz-Garcia, 344 F.3d at 956.

11

Google’s arguments do not identify any such prejudice. It is for the jury to decide how much

12

weight to give to the performance evidence. Google will have the opportunity to contest this

13

evidence at trial and cannot shirk that duty via a motion in limine.

14

IV.

15

For the foregoing reasons, Oracle respectfully asks that the Court deny Google’s motion

CONCLUSION

16

in limine to exclude performance benefits Android gains from use of the claimed invention of the

17

patents-in-suit.

18

Dated: October 4, 2011

19
20
21

MICHAEL A. JACOBS
MARC DAVID PETERS
DANIEL P. MUINO
MORRISON & FOERSTER LLP
By: /s/ Michael A. Jacobs
Michael A. Jacobs

22

Attorneys for Plaintiff
ORACLE AMERICA, INC.

23
24
25
26
27
28

ORACLE’S OPPOSITION TO GOOGLE’S MIL NO. 2 TO EXCLUDE ALLEGED PERFORMANCE BENEFITS
CASE NO. CV 10-03561 WHA
Contains information designated Google Highly Confidential
pa-1488940

14