Beruflich Dokumente
Kultur Dokumente
gov Paper 9
571-272-7822 Entered: March 28, 2017
v.
Case IPR2016-01643
Patent 6,775,745 B1
____________
DECISION
Institution of Inter Partes Review
37 C.F.R. 42.108
I. INTRODUCTION
Unified Patents Inc. (Petitioner) filed a Petition for inter partes
review of claims 14, 69, and 1117 of U.S. Patent No. 6,775,745 B1
(Ex. 1001, the 745 patent). Paper 2 (Pet.). Intellectual Ventures I LLC
(Patent Owner) filed a Preliminary Response. Paper 7 (Prelim. Resp.).
IPR2016-01643
Patent 6,775,745 B1
A. Related Matters
The parties identify the following district court cases in which the
745 patent has been asserted: Intellectual Ventures I, LLC v. NetApp, Inc.,
No. 1:16-cv-10868-IT (D. Mass); and Intellectual Ventures I, LLC v. Lenovo
Grp. Ltd., No. 1:16-cv-10860-IT (D. Mass). Pet. 2; Paper 5, 1.
2
IPR2016-01643
Patent 6,775,745 B1
another small amount for a second program, then seeking back to the
original area to read the next small amount of data for the first program, and
so forth. Id. at 1:4248. The patent explains that to circumvent these
problems, file system drivers do a certain amount of caching. Id. at 2:13.
According to the 745 patent, caching programs strictly use the MRU-LRU
(most recently used-least recently used) mechanism as their sole means of
deciding what data to keep and what data to discard. Id. at 2:1417.
According to the patent, in using the MRU-LRU approach, an important file
that has not been used recently will be discarded when the cache reaches a
target capacity and is forced to free up additional capacity. Id. at 2:1720.
The 745 patent states that [s]trictly adhering to the MRU-LRU mechanism
fails to ensure that files which have been used often, but may not have been
used recently, are maintained in the cache. Id. at 2:2023.
The 745 patent discloses a caching mechanism that minimizes seeks
and reads to a hard drive and which keeps data in the cache based upon
currency of use and the number of times the data is used. Id. at 2:3135.
The patent states that system performance is enhanced through the hybrid
caching mechanism where a file that been used often but not recently used is
maintained in the cache. Id. at 3:4850.
Specifically, the 745 patent discloses identifying a least frequently
used file which has not been recently used and discarding that file when the
cache needs additional capacity to store files. Id. at 4:3944. Figure 2A of
the 745 patent is shown below:
3
IPR2016-01643
Patent 6,775,745 B1
4
IPR2016-01643
Patent 6,775,745 B1
The 745 patent discloses that once the target capacity of the cache is
reached, the frequency factors are scanned, beginning (in one embodiment)
with the frequency factor for the LRU file and ending with the MRU file.
Id. at 6:813. The scanning of the frequency factors determines which
file . . . is the least frequently used (LFU) file which has been least recently
used. Id. at 6:1821. The 745 patent discloses an embodiment in which
the frequency factor for the MRU file is not considered when determining
the LFU file since the MRU file has been so recently placed in the cache and
has not had an opportunity to be reused. Id. at 6:3943. In the example
illustrated in Figure 2A above, the scan will find file F6 134 as the LRU file
with the lowest frequency factori.e., the LFU file which has been least
recently used. Id. at 6:4649. A new file to be placed in the cache, such as
file F9 158, will replace file F6 134. Id. at 6:5255.
The 745 patent describes implementing the improved caching
mechanism in conjunction with reading extending segments of data. Id. at
4:3236. The 745 patent states that large reads (greater than 64 Kbyte)
eliminate the overhead, i.e., seek times, rotational latencies, transfer times,
etc., associated with performing two, four or more reads at 32 Kbytes or less
as performed in the prior art. Id. at 5:3842. For example, the 745 patent
describes that when the first block of the file is read[,] an additional 64,
128, or 256 Kbytes of data are read with it. Id. at 11:2226. The 745
patent states that the reads of the extended segments allows for the
minimization of seeks and reads from the storage medium, since the files are
transferred to cache upon the initial read. Id. at 11:4548.
5
IPR2016-01643
Patent 6,775,745 B1
C. Illustrative Claims
Among the challenged claims (claims 14, 69, and 1117), claims 1,
4, 8, 12, and 15 are independent. Claims 4 and 8 are illustrative of the
subject matter of the challenged claims and read as follows:
4. A caching method for enhancing system performance
of a computer, comprising:
reading an extended segment of data in response to a
request from an operating system;
storing copies of files associated with the extended
segment in a cache;
assigning frequency factors to each of the files stored in
the cache, the frequency factors indicating how often each of the
corresponding files are requested by the operating system;
scanning the frequency factors, the scanning being
performed in response to a target capacity of the cache being
attained;
identifying a least frequently and least recently used file;
and
eliminating the least frequently and least recently used file
to liberate capacity of the cache.
8. A method for caching files in a cache of a random
access memory (RAM) of a computer, comprising:
requesting files, the files being stored on a storage
medium;
reading the files;
copying the files to the cache;
assigning a frequency factor to each of the copied files, the
frequency factor reflecting how often each of the copied files is
accessed;
in response to reaching a target capacity in the cache, the
method further includes,
6
IPR2016-01643
Patent 6,775,745 B1
7
IPR2016-01643
Patent 6,775,745 B1
1
Ramakrishna Karedla et al., Caching Strategies to Improve Disk System
Performance, Computer, Vol. 27, No. 3, 3846 (March 1994) (Ex. 1004,
Karedla).
2
The Leahy-Smith America Invents Act (AIA), Pub. L. No. 112-29, 125
Stat. 284, 28788 (2011), revised 35 U.S.C. 102 and 103, effective March
16, 2013. Because the challenged patent was filed before March 16, 2013,
we refer to the pre-AIA versions of 102 and 103 in this decision.
3
U.S. Patent No. 6,738,865, issued May 18, 2004 (Ex. 1006, Burton).
4
Even though Petitioner identifies claims 1, 4, 69, and 1117 for this
ground in its summary chart, Petitioner does not address claims 7, 13, 14,
and 16 in its analysis of this ground. See Pet. 3, 3956.
5
U.S. Patent No. 5,043,885, issued Aug. 27, 1991 (Ex. 1007, Robinson
885).
6
John T. Robinson & Murthy V. Devarakonda, Data Cache Management
Using Frequency-Based Replacement, Performance Evaluation Rev., Vol.
18, No. 1, 134142 (May 1990) (Ex. 1005, Robinson article).
7
In its summary chart, Petitioner does not identify specifically the Robinson
article as part of this ground. See Pet. 3. In its analysis, however, Petitioner
relies on teachings from the Robinson article. See, e.g., Pet. 57 (relying on
the Robinson FBR algorithm (i.e., Robinson 885 and the Robinson
Article).
8
U.S. Patent No. 6,460,122, issued Oct. 1, 2002 (Ex. 1008, Otterness).
8
IPR2016-01643
Patent 6,775,745 B1
II. DISCUSSION
A. Claim Construction
In an inter partes review, we construe claim terms in an unexpired
patent according to their broadest reasonable construction in light of the
specification of the patent in which they appear. 37 C.F.R. 42.100(b);
Cuozzo Speed Techs., LLC v. Lee, 136 S. Ct. 2131, 214446 (2016)
(upholding the use of the broadest reasonable interpretation standard).
Consistent with the broadest reasonable construction, claim terms are
presumed to have their ordinary and customary meaning as understood by a
person of ordinary skill in the art in the context of the entire patent
disclosure. In re Translogic Tech., Inc., 504 F.3d 1249, 1257 (Fed. Cir.
2007). In addition, the Board may not construe claims during [an inter
partes review] so broadly that its constructions are unreasonable under
general claim construction principles. Microsoft Corp. v. Proxyconn, Inc.,
789 F.3d 1292, 1298 (Fed. Cir. 2015) (emphasis omitted). An inventor may
provide a meaning for a term that is different from its ordinary meaning by
defining the term in the specification with reasonable clarity, deliberateness,
and precision. In re Paulsen, 30 F.3d 1475, 1480 (Fed. Cir. 1994).
Petitioner proposes constructions for the following terms/phrases:
reading extending segment(s) of data; files; frequency factor(s); and
self-tuning. Pet. 1519. Patent Owner asserts that only frequency
factor needs to be construed at this preliminary stage. Prelim. Resp. 59.
For purposes of this decision, we determine that only frequency factor
requires express construction.
The term frequency factor (or frequency factors) is recited in each
challenged claim. Petitioner contends that frequency factor should be
9
IPR2016-01643
Patent 6,775,745 B1
10
IPR2016-01643
Patent 6,775,745 B1
11
IPR2016-01643
Patent 6,775,745 B1
factor reflects how often each of the copied files is accessed (id. at 2:59
61), and that the [t]he frequency factor . . . reflects how frequent each of the
files is accessed (id. at 3:3940). The 745 patent further discloses that, in
one embodiment, [t]he frequency factor is increased with each use of the
file so that the frequency factor indicates how often a file is accessed, and
in another embodiment, the frequency factor is decreased with the lack of
use of the corresponding file. Id. at 5:656:4. The 745 patent also
describes a file with the lowest frequency factor as the file that is least
frequently used. See, e.g., id. at 6:4649. The 745 patent also uses the
terms frequency factor, frequency of use, frequency of use factor,
frequency of use count, and frequency count interchangeably. See id. at
10:125. Moreover, in describing the flow chart of Figure 4, the 745 patent
equates frequencies shown in the figure with frequency factors. Id. at
10:4446, 10:5455.
The 745 patent, however, also describes scenarios for weighting or
decrementing frequency factors based on criteria other than frequency of use
or access. See id. at 6:567:15, 8:659:3. For example, the 745 patent
describes an embodiment in which a user finishes watching a movie and
begins word processing operations, but the cache has files from the movie,
which were frequently used. Id. at 6:5760. The 745 patent describes
decrementing the frequency factor associated with the movie files to bring
down the frequency factor so that the files are removed from the cache. Id.
at 6:6366. The patent also describes weighting certain files with
frequency factors on their initial read into the cache to guarantee they remain
above others. Id. at 7:311; see also id. at 8:659:3 (disclosing that the
frequency factor would be incremented by a factor of more than one for
12
IPR2016-01643
Patent 6,775,745 B1
each time the directory or FAT data is accessed). Thus, the specification
describes frequency factors that are based in part on frequency of use and
based in part on other criteria, such as nature of the data accessed.
Dependent claim 3 also recites a limitation directed to decrementing the
frequency factor based on criteria other than frequency of access. Id. at
12:5154 (claim 3 reciting providing a driver for decrementing a frequency
factor after a time period of non-use of a corresponding file).
Accordingly, based on the current record, the specification
considered in its entiretydoes not limit frequency factor to a one-to-one
correspondence with frequency of use or access, even though many
examples of a frequency factor are that it has a value that is the same as the
number of times a corresponding file is used or accessed. We, therefore,
determine on the current record and at this preliminary stage that Petitioners
proposed constructionthat the frequency factor is an indicator for
distinguishing data based in part on its frequency of useis too narrow in
that may exclude a frequency factor that is based entirely on frequency of
use. We also determine that Patent Owners proposed constructionan
indication of how often a file is requested or access by the operating
systemmay be read as too narrow in that it potentially may exclude a
frequency factor that is based only in part on frequency of use. Patent
Owners position that the frequency factor cannot be only tangentially
related to frequency of use is too vague and amorphous on the current
record. The general term frequency factor does not dictate a particular
formula for how much frequency must be weighed. The 745 patent
discloses generally that the frequency factor may be decreased or increased
because of, for example, expected non-use or because the data is directory
13
IPR2016-01643
Patent 6,775,745 B1
data. Id. at 6:637:3, 8:659.3. The specification but does not limit how
much the frequency factor is affected by such criteria. Even so, given the
use of the word frequency in the claim term, the frequency factor must be
based to some extent on frequency of access or use.
For the above reasons, and for purposes of institution and on the
current record, we determine that the broadest reasonable interpretation of
frequency factor is an indicator based on frequency of use or access.
1. Summary of Karedla
Karedla is an article from the March 1994 issue of Computer titled
Caching Strategies to Improve Disk System Performance. Ex. 1004, 1, 6.9
The cover page and table of contents of Karedla are dated March 1994 and
the second page of Karedla has a stamp of the Copyright Office of the
Library of Congress dated March 22, 1994. Id. at 1, 2, 4.
Karedla characterizes its disclosure as follows:
In this article, we examine the use of caching as a means to
increase system response time and improve the data throughput
of the disk subsystem. After explaining some basic caching
issues . . . , we examine some popular caching strategies and
9
Even though Petitioner cites to the original page numbers of Karedla rather
than the exhibit page numbers added by Petitioner, for ease of reference to
the entire exhibit, we cite to the exhibit page numbers.
14
IPR2016-01643
Patent 6,775,745 B1
15
IPR2016-01643
Patent 6,775,745 B1
amount of data that is prefetched. Id. Karedla discloses that [a] read-
ahead strategy known as prefetching exploits the principle of spatial locality
in an attempt to minimize latency in data access by anticipating future
requests for data and bringing it into the cache. Id.
Karedla further discloses that the authors simulated the behavior of
three cache replacement algorithms: random replacement, least recently
used, and segmented LRU. Id. at 1011. According to Karedla, the
simulated segmented LRU algorithm is the simplest variation on frequency-
based LRU replacement algorithms; it modif[ies] the LRU replacement
algorithm to distinguish between lines that have been accessed only once
and lines that have been accessed multiple times. Id. at 11.
Figure 3 of Karedla is shown below:
Id. Figure 3 above depicts the logical flow of SLRU cache lines between a
protected segment and a probationary segment. Id. Karedla describes the
segmented LRU cache as follows:
16
IPR2016-01643
Patent 6,775,745 B1
2. Analysis
17
IPR2016-01643
Patent 6,775,745 B1
b. Claim 4
Petitioner contends that Karedla discloses each limitation of
independent claim 4. Id. at 2028. For example, Petitioner asserts that
Karedla discloses the claimed reading an extended segment of data in
response to a request from an operating system by describing the read-
ahead strategy known as prefetching, which anticipat[es] future requests
for data and bring[s] it into the cache. Id. at 21 (quoting Ex. 1004, 8).
Petitioner also relies on Karedlas descriptions of the LFU algorithm
maintaining a frequency-of-access count for all lines in the cache and of the
segmented LRU algorithm discarding data from the LRU end of the
probationary segment (which, according to Petitioner, is the least frequently
and least recently used file, as claimed). Id. at 2328.
Patent Owner asserts that Petitioner improperly relies on separate
embodiments described in Karedla for meeting the reading extended data
limitation and thus does not show that Karedla discloses all of the limitations
as arranged in the claim. Prelim. Resp. 1114. Patent Owner also argues
that Karedlas description of the variations on LRU embodiment does not
disclose assigning frequency factors to each of the files stored in the
cache, as claimed, and Petitioners reliance on a separate referencethe
Robinson articleto show this for anticipation by Karedla is improper. Id.
at 1518. Patent Owner further argues that Karedlas description of the
segmented LRU embodiment also does not disclose assigning frequency
factors to each of the files stored in the cache, as claimed. Id. at 1820.
Based on the current record, the information presented establishes a
reasonable likelihood that Petitioner would prevail in showing that claim 4 is
unpatentable under 35 U.S.C. 102(b) as anticipated by Karedla. See Pet.
18
IPR2016-01643
Patent 6,775,745 B1
19
IPR2016-01643
Patent 6,775,745 B1
c. Claims 1 and 12
Petitioner asserts that claims 1 and 12 are substantively identical to
claim 4, noting that claims 1 and 12 are respectively directed to a similar
method and media for identifying a least frequently and least recently used
file. Pet. 28. Petitioner asserts that Karedla discloses the feature recited in
claims 1 and 12, but not in claim 4, of the cache being located in a random
access memory. Pet. 2831 (citing, e.g., Ex. 1004, 7, 9; Ex. 1002 104,
107). Patent Owner raises the arguments discussed above in connection
with claim 4 also for claims 1 and 12. See Prelim. Resp. 920.
Claim 1 is a method claim that recites limitations that are very similar
to those of method claim 4, with the additional requirement of the cache
being located in a random access memory (RAM) of the computer. Claim
12 recites [a] computer readable media having program instructions for a
caching method comprising program instructions for performing functions
nearly identical to those recited in claim 4. Like claim 1, claim 12
additionally requires that the cache be located in a random access memory of
the computer.
For the reasons explained above in connection with claim 4, and
because Petitioner sufficiently shows for institution purposes that Karedla
discloses that the cache is located in a random access memory, the
information presented establishes a reasonable likelihood that Petitioner
would prevail in showing that claims 1 and 12 are unpatentable under
35 U.S.C. 102(b) as anticipated by Karedla. We also are satisfied, for this
ground, that Petitioners citations to Karedla are sufficient for institution
purposes and do not rely on any of Petitioners citations to the Robinson
article. See, e.g., Pet. 29 (citing Ex. 1005).
20
IPR2016-01643
Patent 6,775,745 B1
d. Claims 8, 9, and 11
Petitioner asserts that claim 8 is substantively identical to claim 4.
Pet. 28. Petitioner, however, notes that, unlike claim 4, claim 8 requires
scanning the frequency factor . . . without considering a frequency factor
corresponding to a most recently used file. Id. at 31; see Ex. 1001, 13:26
30. Petitioner asserts that Karedla discloses this limitation because
Karedlas SLRU variation does not consider the frequency factor
corresponding to the MRU file. Pet. 33 (citing Ex. 1002 116).
Specifically, Petitioner asserts that [b]ecause the MRU end of the protected
segment is protected and only the LRU probationary entry is chosen for
discarding, the SLRU algorithm . . . does not consider the frequency factor
corresponding to the MRU file in the protected segment. Pet. 33 (quoting
Ex. 1004, 11). Petitioner also cites to the Robinson article and Robinson
885 for descriptions of the Robinson FBR algorithm, which Petitioner
asserts also does not consider the frequency factor corresponding to the
MRU file. Id. at 3133.
Patent Owner asserts that Karedla does not disclose scanning the
frequency factor for each of the files, as recited in claim 8. Prelim. Resp.
20. Patent Owner specifically asserts that the scanning limitation of claim 8
expressly requires scanning the frequency factor of every file in the cache
to determine which files to replace in the cache, and that [b]oth Karedla
embodiments scan only a subset of cache lines to determine which cache
lines to replace in the cache. Id. With respect to the variations on LRU
embodiment, Patent Owner asserts that an ordinarily skilled artisan would
understand that Karedlas disclosure that a cache line that is not
subsequently referenced ages into the bottom part of the list by LRU
21
IPR2016-01643
Patent 6,775,745 B1
replacement and is finally evicted if not referenced to mean that only the
cache lines in the bottom section of the stack are scanned to determine which
cache lines to replace. Id. at 2021. With respect to the segmented LRU
embodiment, Patent Owner asserts that only the probationary segment is
scanned to determine which cache lines to replace. Id. at 21.
Having reviewed the record, we determine that Petitioner does not
show sufficiently that Karedla discloses scanning the frequency factor for
each of the files, as required by claim 8. Petitioner relies on Karedlas
disclosures that a cache line, after being referenced, is placed in the
topmost section of the stack and that the line ages into the bottom part of
the list . . . and is finally evicted if not referenced (Pet. 32 (citing Ex. 1004,
8)); that the SLRU linked list runs from the MRU end of the protected
segment to the LRU end of the probationary segment (id. at 33 (citing Ex.
1004, 11)); and that the LRU probationary entry is chosen for discarding
(id. (citing Ex. 1004, 11)). These excerpts from Karedla, however, do not
describe scanning the frequency factor for each of the files. Petitioner fails
to address this specific requirement of the claim or explain how Karedla
expressly or inherently discloses this limitation. See Pet. 3133. Indeed,
Petitioner does not even quote this part of the claim language when
explaining how claim 8 differs from claim 4. Id. at 31.
We, therefore, determine that the information presented does not show
a reasonable likelihood that Petitioner would prevail in demonstrating that
claim 8 is anticipated by Karedla, or that claims 9 and 11, which depend
from claim 8, are anticipated by Karedla.
22
IPR2016-01643
Patent 6,775,745 B1
e. Claims 1517
Petitioner asserts that claim 15 is substantively identical to claim 4.
Pet. 28. Petitioner, however, notes that, unlike claim 4, claim 15 requires
that the frequency factors are scanned from a frequency factor for a LRU
file to a frequency factor for a MRU file, not including the frequency factor
for the MRU file. Pet. 3334; see Ex. 1001, 14:3942.
Petitioner relies on its analysis of the scanning limitation of claim 4
and asserts that a person of ordinary skill in the art would understand that
Karedla discloses scanned from a frequency factor for a LRU file to a
frequency factor for a MRU file, not including the frequency factor for the
MRU file. Id. at 3334.
Patent Owner argues that claim 15 requires scanning every frequency
factor except the frequency factor for the MRU file and asserts that
Karedlas variations on LRU embodiment scans only the cache lines in the
bottom section of a three-segment stack and the segmented LRU
embodiment scans only the cache lines in the probationary segment of a
two-segment cache. Prelim. Resp. 2122.
Having reviewed the record, we determine that Petitioner does not
show sufficiently that Karedla discloses the frequency factors are scanned
from a frequency factor for a LRU file to a frequency factor for a MRU file,
not including the frequency factor for the MRU file, as required by claim
15. Petitioner does not rely on any express disclosure in Karedla. Rather,
Petitioner relies on a disclosure in Robinson 885 that [f]inding a block to
replace then consists of scanning the blocks (from LRU to MRU) in each
such LRU chain (in ascending count order) until a block is found in the old
section. Pet. 34 (quoting Ex. 1007, 5:1518). Petitioner then asserts that a
23
IPR2016-01643
Patent 6,775,745 B1
[person of ordinary skill in the art] reading Karedla would understand that
Karedla discloses scanned from a frequency factor for a LRU file to a
frequency factor for a MRU file, not including the frequency factor for the
MRU file. Id. (citing Ex. 1002 119) (emphasis omitted). The cited
declaration testimony likewise relies on the quoted disclosure from
Robinson 885 and repeats the assertion that a skilled artisan would
understand that Karedla discloses the claim limitation. See Ex. 1002 119
(emphasis omitted). Karedla, however, does not cite Robinson 885, let
alone incorporate by reference this particular disclosure from Robinson
885. See, e.g., Ex. 1004, 8, 14. Petitioner, thus, fails to show that Karedla
expressly or inherently discloses this claim limitation.
For these reasons, we determine that the information presented does
not show a reasonable likelihood that Petitioner would prevail in
demonstrating that claim 15 is anticipated by Karedla, or that claims 16 and
17, which depend from claim 15, are anticipated by Karedla.
f. Dependent claims 3, 7, and 13
Claim 3 recites that the method recited in claim 1 further includes
providing a driver for decrementing a frequency factor after a time period
of non-use of a corresponding file. Petitioner asserts that a person of
ordinary skill in the art reading Karedla would understand that the
frequency factors may be decremented periodically in the Robinson FBR
algorithm, as described in both the Robinson Article and Robinson 885.
Pet. 36 (citing Ex. 1007, Ex. 1005) (emphasis omitted). Petitioner does not
rely on any express disclosure in Karedla. Id. We determine that Petitioner
fails to show sufficiently that Karedla itself, which does not incorporate by
reference the relied-upon disclosures from the Robinson article or Robinson
24
IPR2016-01643
Patent 6,775,745 B1
885, discloses this claim limitation. See, e.g., Ex. 1004, 40, 46 n.4. We,
therefore, determine that the information presented does not show a
reasonable likelihood that Petitioner would prevail in demonstrating that
claim 3 is unpatentable under 35 U.S.C. 102(b) as anticipated by Karedla.
Claim 7 depends from claim 4 and recites wherein the caching
method is self tuning. Claim 13 depends from claim 12 and similarly
recites that program instructions for the caching method are self-tuning.
Petitioner asserts that Karedla discloses the self-tuning limitations of claims
7 and 13. Pet. 3637. Petitioner also relies on disclosures from the
Robinson article and Robinson 885 that purportedly disclose the claimed
self-tuning. Id. at 36. Patent Owner does not raise any arguments specific to
the limitations added by dependent claims 7 and 13 in its Preliminary
Response. See Prelim. Resp. 922. The burden, however, remains on
Petitioner to demonstrate unpatentability. See Dynamic Drinkware, LLC v.
Natl Graphics, Inc., 800 F.3d 1375, 1378 (Fed. Cir. 2015) (discussing the
burden of proof in inter partes review).
Having reviewed the record, we determine that the information
presented does not establish a reasonable likelihood that Petitioner would
prevail in showing that claims 7 and 13 are unpatentable under 35 U.S.C.
102(b) as anticipated by Karedla. Petitioner asserts that Karedla discloses
that the cache line sizes could be tunable and that one such cache
dynamically partitions the cache into three partitions for small, medium and
large transfers, depending on the request size distribution. Pet. 3637
(quoting Ex. 1004, 8). Petitioner also asserts that Karedla discloses
adaptively sizing the cache. Id. at 37 (quoting Ex. 1004, 9). Petitioner also
relies on the last passage of Karedla, which notes that future I/O caches will
25
IPR2016-01643
Patent 6,775,745 B1
10
Even though Petitioner cites to the original page numbers of the Robinson
article rather than the exhibit page numbers added by Petitioner, for ease of
reference to the entire exhibit, we cite to the exhibit page numbers.
26
IPR2016-01643
Patent 6,775,745 B1
article). We, thus, determine that Petitioner does not show sufficiently that
Karedla discloses the self-tuning recited in claims 7 and 13.
g. Dependent claims 6 and 14
Claim 6 depends from claim 4 and recites wherein the scanning the
frequency factors further includes: scanning from a frequency factor
corresponding to a LRU file to a frequency factor corresponding to a MRU
file. Petitioner relies on its analysis of claim 15 for showing Karedla
anticipates claim 6. Pet. 37. For the reasons explained above in connection
with claim 15, we determine that the information presented does not show a
reasonable likelihood that Petitioner would prevail in demonstrating that
claim 6 is unpatentable under 35 U.S.C. 102(b) as anticipated by Karedla.
Claim 14 recites that the computer readable media as recited in claim
12 further comprises program instructions for allowing a user to adjust a
size of the extended segment that is read. Petitioner relies on its analysis of
claim 4 and additionally relies, for example, on a passage in Karedlas
description of read-ahead strategies stating that vendors also offer a tunable
read-ahead threshold to minimize both cache pollution and the latency in
transferring large amounts of data from the disk to the I/O bus. Pet. 3839
(citing Ex. 1004, 8). Patent Owner does not raise, in its Preliminary
Response, any arguments specific to the limitation added by dependent
claim 14. See Prelim. Resp. 922. The burden, however, remains on
Petitioner to demonstrate unpatentability. See Dynamic Drinkware, 800
F.3d at 1378. Based on the current record, the information presented
establishes a reasonable likelihood that Petitioner would prevail in showing
that claim 14 are unpatentable under 35 U.S.C. 102(b) as anticipated
Karedla. See Pet. 3639.
27
IPR2016-01643
Patent 6,775,745 B1
h. Summary
For the above reasons, we determine that the information presented
(i) establishes a reasonable likelihood that Petitioner would prevail in
showing that claims 1, 4, 12, and 14 are unpatentable under 35 U.S.C.
102(b) as anticipated by Karedla alone (and not including Robinson 885
or the Robinson article) and (ii) does not establish a reasonable likelihood
that Petitioner would prevail in showing that claims 3, 69, 11, 13, and 15
17 are unpatentable under 35 U.S.C. 102(b) as anticipated Karedla.
1. Summary of Burton
Burton is a U.S. patent titled Method, System, and Program for
Demoting Data from Cache Based on Least Recently Accessed and Least
Frequently Accessed Data. Ex. 1006, [54]. Burton issued May 18, 2004
from an application filed June 9, 2000, and is prior art to the 745 under
35 U.S.C. 102(e). Compare id. [22], [45], with Ex. 1001, [22], [45].
Burton describes as background a least recently used (LRU) algorithm
for managing data in a cache. Ex. 1006, 1:3949. Burton explains that
when a track is added to the cache, a pointer to the track is placed at the top
of the LRU linked list. Id. at 1:3942. Burton explains that when the cache
28
IPR2016-01643
Patent 6,775,745 B1
29
IPR2016-01643
Patent 6,775,745 B1
30
IPR2016-01643
Patent 6,775,745 B1
Burton discloses that the LRU rank provides a value for each entry in
[the cache] that indicates both how frequently the entry is accessed and the
time of last access. Id. at 3:1113. Burton states that [t]he purpose of this
weighting is to allow the process [of] demoting entries from cache to take
into account the frequency with which an entry was accessed and remove
those less frequently accessed entries. Id. at 3:1317 (reference numerals
omitted). Burton describes its ranking methodology, which uses time
counter 18 (shown in Figure 1 above), as follows:
The . . . LRU rank is weighted for previous accesses. Because
the time counter 18 is incremented for every I/O access, the value
added to the LRU rank for subsequently accessed cache entries
increases substantially in a short period of time. Thus, LRU
entries recently accessed or added to cache will have a
substantially higher LRU rank than the LRU rank of entries that
have not been accessed recently, even those entries accessed
numerous times. However, to the extent entries have not been
accessed for the same amount of time, the entry that was accessed
more frequently will have a higher LRU rank because its LRU
rank will be weighted with previous accesses. Thus, to the extent
entries were last accessed at about the same time, which in terms
of cache operations is within fractions of a second of each other,
the more frequently accessed entry will have a greater weighting.
Id. at 3:544:2. Burton explains that the more frequently accessed entry is
given a greater weighting because a cache entry more frequently accessed
in the past is likelier to be accessed in the future over less frequently
accessed entries that were last accessed at about the same time. Id. at 4:2
5.
Burton also discloses that when the number of cached entries reaches
an upper threshold, the cache determines from the last 1024 entries in the
LRU linked list thirty-two entries that have the lowest LRU rank. Id. at 4:7
31
IPR2016-01643
Patent 6,775,745 B1
11. According to Burton, the cache then demotes those thirty-two entries
from the cache. Id. at 4:1113. Burton states that those entries having the
lowest LRU rank marked for demotion are both the least recently accessed
and[,] among those entries least recently accessed, are less frequently
accessed. Id. at 4:1922. Thus, according to Burton, the algorithm . . .
ensures that more frequently accessed cache entries . . . remain in cache
longer. Id. at 4:2224.
2. Analysis
a. Claim 4
Petitioner contends that claim 4 of the 745 patent is unpatentable
under 35 U.S.C. 103(a) as obvious over Burton and Karedla. Pet. 3, 40
50. Relying in part on the testimony of Dr. Franzon, Petitioner explains how
the references allegedly teach or suggest the claim limitations and provides
purported reasoning for combining the teachings of the references. Id. at
4050 (citing Ex. 1002 134, 135, 141, 142, 144148, 150). Specifically,
Petitioner asserts that Burton discloses all of the limitations of claim 4,
except reading an extended segment of data in response to a request from
an operating system, which Petitioner asserts Karedla discloses. Id. at 40
49. Petitioner also asserts that a person of ordinary skill in the art would
have been motivated to combine Burtons cache replacement mechanism
with Karedlas cache strategies to improve system performance, and that the
combination provides a cost-effective method to increase performance and
data throughput, minimize latency in data access, reduce cache lookup time,
and increase the speed of cache directory searches. Id. at 4950 (citing Ex.
1002 158; Ex. 1006, 1:6162; Ex. 1004, 8).
32
IPR2016-01643
Patent 6,775,745 B1
33
IPR2016-01643
Patent 6,775,745 B1
34
IPR2016-01643
Patent 6,775,745 B1
Id. at 5:1318 (emphasis added). Thus, although Patent Owner relies on the
specific algorithm shown in Figure 2 for distinguishing Burton, Burton
discloses that other algorithms may be used.
Petitioner also sufficiently shows that Burton discloses the other
limitations of claim 4, except for reading extended segments of data,
which Petitioner sufficiently shows Karedla discloses. See Pet. 3949.
Patent Owner again argues that Petitioner does not establish that Karedla
discloses reading extended segments of data as claimed because Petitioner
improperly relies on different embodiments in Karedla. Prelim. Resp. 30
31. As explained above in connection with the anticipation ground based on
Karedla, at this preliminary stage, Petitioner sufficiently shows that Karedla
discloses the claimed reading extended segments of data.
Petitioner also provides, on the current record and for institution
purposes, sufficient reasoning with rational underpinning for combining the
teachings of Burton and Karedla as claimedto improve system
performance such as to minimize latency in data excess, to reduce cache
lookup time, and to increase the speed of cache directory searches. See Pet.
4950.
For the above reasons, the information presented shows a reasonable
likelihood that Petitioner would prevail in demonstrating that claim 4 is
unpatentable for obviousness over Burton and Karedla.
b. Claims 1 and 12
Petitioner asserts that claims 1 and 12 are substantively identical to
claim 4, noting that claims 1 and 12 are respectively directed to a similar
method and media for identifying a least frequently and least recently used
file. Pet. 50. Petitioner also asserts that Burton discloses the feature recited
35
IPR2016-01643
Patent 6,775,745 B1
36
IPR2016-01643
Patent 6,775,745 B1
the least frequently used and least recently used files from the bottom of the
LRU linked list, and (iii) that the thirty-two entries that have the lowest LRU
rank from the last 1024 entries are identified. Id. (citing Ex. 1006, 3:67,
3:5861, 4:911; Ex. 1002 162, 163).
Patent Owner does not specifically address Petitioners contention that
the subject matter of claim 8 is rendered obvious by the combination of
Burton and Karedla, but does address Petitioners contention that the subject
matter of claims 9 and 11which depend from claim 8would have been
obvious in view of Burton and Karedla. See Prelim. Resp. 3132. Patent
Owner asserts that the scanning limitation of claim 8 expressly requires
scanning the frequency factor of every file in the cache to determine which
files to replace in the cache, and that Burton expressly discloses that it
scans only a subset of cache entries to determine which cache entries to
replace in the cache. Id. at 31. In addition, in response to Petitioners
contention that claim 8 is anticipated by Burton, Patent Owner asserts that
Burton does not disclose assigning a frequency factor to each of the copied
files or scanning the frequency factor for each of the files, as recited in
claim 8. Prelim. Resp. 34.
Having reviewed the record, we determine that Petitioner fails to
show sufficiently that the subject matter of claim 8 would have been obvious
in view of Burton and Karedla. Petitioner fails to address how Burton
discloses the claim requirement of scanning the frequency factor for each
of the files, which is not recited in independent claim 4. See Pet. 52. Burton
states that the cache determines . . . from the last 1024 entries in the LRU
linked list 8, thirty-two entries that have the lowest LRU rank 12a, b, c, d.
The cache 2 then demotes . . . those determined thirty-two entries from
37
IPR2016-01643
Patent 6,775,745 B1
cache 2. Ex. 1006, 4:713. Petitioner fails to show, in view of this express
disclosure in Burton, how Burton teaches scanning the frequency factor for
each of the files in the cache. Petitioner does not rely on any disclosure in
Karedla for meeting this claim limitation of scanning the frequency factor
for each of the files. See Pet. 52. We, therefore, determine that the
information presented does not show a reasonable likelihood that Petitioner
would prevail in demonstrating that the subject matter of claim 8 would have
been obvious over Burton and Karedla.
d. Claim 15
Petitioner asserts that claim 15 is substantively identical to claim 4.
Pet. 50. Petitioner however notes that, unlike claim 4, claim 15 requires that
the frequency factors are scanned from a frequency factor for a LRU file to
a frequency factor for a MRU file, not including the frequency factor for the
MRU file. Id. at 53; see also Ex. 1001, 14:3942. Petitioner asserts that
Burton teaches this limitation in that (i) Burton discloses that the top of the
LRU linked list is the MRU entry and the bottom of the LRU linked list is
the LRU entry, (ii) Burton discloses that the bottom section of the LRU
linked list is scanned for the least recently accessed and least frequently
accessed data, (iii) a person of ordinary skill in the art reviewing Burton
would understand that the frequency factor of the MRU file is not scanned in
Burtons scanning process because only the bottom portion is scanned,
(iv) because Burton discloses scanning the bottom section of the LRU linked
list, the scanning is neither haphazard nor from the MRU to the LRU, and
(v) a skilled artisan reading Burton would understand that the scanning
process necessarily must scan from the MRU to the LRU, and there would
38
IPR2016-01643
Patent 6,775,745 B1
be no way for the system to identify this block without such scanning. Id.
at 54 (citing, e.g., Ex. 1002 165, 167).
Patent Owner argues that claim 15 expressly requires scanning the
frequency factor of every file in the cache, except the MRU file, to
determine which files to replace in the cache and that Burton does not
disclose this limitation because Burton expressly discloses that only a subset
of cache entries are scanned to determine which cache entries to replace in
the cache. Prelim. Resp. 3233 (bold emphasis omitted).
Having reviewed the record, we determine that Petitioner fails to
show sufficiently that the subject matter of claim 15 would have been
obvious in view of Burton and Karedla. Burton expressly discloses scanning
a bottom section of the cache, but does not disclose scanning the frequency
factors from a frequency factor for a LRU file to a frequency factor for a
MRU file, not including the frequency factor for the MRU file, as required
by claim 15. See Ex. 1006, 4:813. Petitioner relies on testimony of its
expert to purportedly show that Burton inherently discloses this limitation
that the scanning necessarily must scan from the MRU to the LRU, and
there would be no way for the system to identify this block without such
scanning. Pet. 54 (citing Ex. 1002 167). Neither Petitioner nor Dr.
Franzon, however, explains why the scan must be from the MRU file to the
LRU file or why there would be no way for the system to identify the bottom
block without scanning all of the files. See id.; Ex. 1002 167. Given
Burtons express disclosure of scanning a particular subsection of the cache,
Petitioners lack of explanation in support of its inherency argument is
insufficient. Petitioner does not rely on any disclosure in Karedla for
disclosing this limitation. Pet. 54. For these reasons, we determine that the
39
IPR2016-01643
Patent 6,775,745 B1
40
IPR2016-01643
Patent 6,775,745 B1
41
IPR2016-01643
Patent 6,775,745 B1
42
IPR2016-01643
Patent 6,775,745 B1
43
IPR2016-01643
Patent 6,775,745 B1
As shown in Figure 1, the cache directory has age boundary 18. The
section of the cache directory on the MRU side of the boundary is the
local section; the section on the other side of the boundary is the non-
local section. Id. at 4:4647. Robinson 885 discloses that when there is a
hit on the local section, the count remains the same; when there is a hit on
the non-local section, the count is incremented. Id. at 4:4852. Robinson
885 discloses ways of using the reference counts to select blocks to replace,
including select[ing] the least recently used block in the non-local section
whose count is below a preselected threshold. Id. at 4:5557.
Figure 2 of Robinson 885, shown below, illustrates an alternative
embodiment in which the cache directory is divided into three sections using
two boundaries:
44
IPR2016-01643
Patent 6,775,745 B1
Id. at 11. As illustrated in Figure 2.2 above, the cache includes a new
section, a middle section, and an old section. Id. The Robinson article
45
IPR2016-01643
Patent 6,775,745 B1
discloses that the basis of its algorithm is the LRU replacement algorithm, in
which the cache consists of a stack of blocks, with the most recently
referenced block always pushed to the top of the stack. Id. at 10. The
Robinson article discloses that a certain portion of the top part of the stack is
set as aside as the new section, a bottom part of the stack is defined as the
old section, and the remaining part of the stack between the new and old
sections is the middle section. Id. at 11. According to the Robinson article,
replacement choices are confined to those blocks that have aged into an
old section of the cache. Id. at 10.
3. Analysis
a. Claim 3
Claim 3 depends from claim 1 and further recites providing a driver
for decrementing a frequency factor after a time period of non-use of a
corresponding file. Petitioner asserts that Robinson 885 discloses this
limitation in that it discloses that in order to adapt to changes in the
underlying frequency distribution, all counts periodically may be
decremented or multiplied by a reduction factor. Pet. 61 (quoting Ex. 1007,
9:712), 64 (same). Petitioner also cites to Karedlas disclosure of
periodically adjusting reference counts so as to approximate reference
frequencies and avoid practical problems. Pet. 61 (quoting Ex. 1004, 135),
64 (same). Despite these citations, Petitioner does not show sufficiently that
either Karedla or Robinson 885 discloses decrementing a frequency factor
after a time period of non-use of a corresponding file, as claimed. Nor
does Petitioner rely on the Robinson article for disclosing this feature.
Additionally, Petitioner fails to demonstrate that the combined teachings of
the cited references would have taught the challenged limitation to one of
46
IPR2016-01643
Patent 6,775,745 B1
ordinary skill in the art. We, therefore, determine that the information
presented does not show a reasonable likelihood that Petitioner would
prevail in demonstrating that subject matter of claim 3 would have been
obvious over Karedla, Robinson 885, and the Robinson article.
b. Claims 6, 15, and 16
Claim 6 requires scanning from a frequency factor corresponding to
a LRU file to a frequency factor corresponding to a MRU file. Claim 15
similarly requires the frequency factors are scanned from a frequency factor
for a LRU file to a frequency factor for a MRU file, not including a
frequency factor for the MRU file. As explained in Sections II.C.2.e and
II.C.2.g above, we determine that the information presented does not
establish a reasonable likelihood that Petitioner would prevail in showing
that claims 6 and 15 are unpatentable under 35 U.S.C. 102(b) as
anticipated by Karedla. Petitioner further asserts that the subject matter of
claims 6 and 15 would have been obvious in view of Karedla, Robinson
885, and the Robinson article. Pet. 57, 6466. Specifically, Petitioner
asserts that Robinson 885 discloses the claimed scanning by disclosing that
[f]inding a block to replace . . . consists of scanning the blocks (from LRU
to MRU) in each such LRU chain (in ascending count order) until a block is
found in the old section. Pet. 64, 66 (quoting Ex. 1007, 5:1518); see also
Ex. 1002 119 (same).
Having reviewed the record, we determine that the information
presented does not establish a reasonable likelihood that Petitioner would
prevail in showing that claims 6 and 15 are unpatentable under 35 U.S.C.
103(a) as obvious over Karedla, Robinson 885, and the Robinson article.
Petitioner fails to explain how the quoted passage from Robinson 885
47
IPR2016-01643
Patent 6,775,745 B1
48
IPR2016-01643
Patent 6,775,745 B1
Petitioners declarant simply repeats the statements from the Petition without
providing any further explanation, and thus, fails to help Petitioner carry its
burden. See Ex. 1002 184, 188. Petitioner also does not explain
sufficiently how the combination of Karedla, Robinson 885, and the
Robinson article teach this feature. See Pet. 5766. Accordingly, we
determine that the information presented does not establish a reasonable
likelihood that Petitioner would prevail in showing that the subject matter of
claims 6 or 15, or claim 16, which depends from claim 15, would have been
obvious over Karedla, Robinson 885, and the Robinson article.
c. Claims 7 and 13
Claim 7 depends from claim 4 and recites wherein the caching
method is self tuning. Claim 13 depends from claim 12 and recites that
program instructions for the caching method are self-tuning. Petitioner
asserts that Karedla discloses the self-tuning limitations as explained in
connection with the anticipation ground and asserts that Robinson 885 and
the Robinson article expressly teach[] a more complete description of these
elements. Pet. 57. Specifically, Petitioner cites to the disclosure in
Robinson 885 that in order to adapt to changes in the underlying frequency
distribution, all counts periodically may be decremented or multiplied by a
reduction factor. Id. at 61 (quoting Ex. 1007, 9:712); see also Ex. 1002
188 (same).
In its Preliminary Response, Patent Owner does not raise any
arguments specific to the limitations added by dependent claims 7 and 13.
See Prelim. Resp. 3435. The burden, however, remains on Petitioner to
demonstrate unpatentability. See Dynamic Drinkware, 800 F.3d at 1378.
49
IPR2016-01643
Patent 6,775,745 B1
50
IPR2016-01643
Patent 6,775,745 B1
1. Summary of Otterness
Otterness is a U.S. patent that issued October 1, 2002 from an
application filed September 30, 1999, and is prior art to the 745 patent
under 35 U.S.C. 102(e). Compare Ex. 1008, [22], [45], with Ex. 1001,
[22], [45]. Otterness discloses a multiple level caching method that
distributes caching operations across multiple processors to provide higher
input/output processing performance. Ex. 1008, 1:1017. Otterness
discloses using a cache line descriptor (CLD) to keep track of all of the data
stored in the cache. Id. at 9:2325.
Figure 5 of Otterness is shown below:
51
IPR2016-01643
Patent 6,775,745 B1
2. Analysis
Claim 2 recites [t]he method as recited in claim 1, wherein the
extended segments are one of 64 Kbytes, 128 Kbytes and 256 Kbytes in
size. Petitioner asserts that a person of ordinary skill in the art reading
Otterness would have understood that, although Otterness does not
expressly list out cache line sizes of 128 kbyte or 256 kbyte, a skilled
52
IPR2016-01643
Patent 6,775,745 B1
53
IPR2016-01643
Patent 6,775,745 B1
54
IPR2016-01643
Patent 6,775,745 B1
did not file a motion seeking to waive the word count. Id. at 41. Patent
Owner argues the Board should exercise its discretion to deny the Petition
or to impose another appropriate remedy. Id. at 42.
We agree with Patent Owner that Petitioner uses an atypical citation
format for its exhibit cites in that the common practice is to cite an exhibit
using a space separating the abbreviation for exhibit from the exhibit
number. We are, however, unaware of a requirement that an exhibit be cited
using two separate wordsone for an abbreviation of the word exhibit
and one for the exhibit number. We, therefore, decline to exercise our
discretion to deny the Petition, as requested by Patent Owner. We, however,
recognize a potential advantage of being able to use more words in a
submission by using this atypical exhibit citation practice. Accordingly, we
note that Patent Owner, in its Patent Owner Response in the inter partes
review instituted herein, may use a similar atypical format for citing to
exhibits. Moreover, we encourage Petitioner to use the common practice for
citing exhibits in its Reply and any other papers subsequently filed in the
proceeding.
III. CONCLUSION
For the above reasons, we determine that the information presented
establishes a reasonable likelihood that Petitioner would prevail in showing
that (i) claims 1, 4, 12, and 14 are unpatentable under 35 U.S.C. 102(b) as
anticipated Karedla, (ii) claims 1, 4, and 12 are unpatentable under
35 U.S.C. 103(a) as obvious over Burton and Karedla, and (iii) claim 2 is
unpatentable under 35 U.S.C. 103(a) as obvious over either Karedla and
Otterness or Burton, Karedla, and Otterness. We also determine that the
information presented does not show a reasonable likelihood that Petitioner
55
IPR2016-01643
Patent 6,775,745 B1
IV. ORDER
Accordingly, it is:
ORDERED that pursuant to 35 U.S.C. 314(a), an inter partes
review is instituted as to (i) claims 1, 4, 12, and 14 of the 745 patent on the
ground of anticipation by Karedla, (ii) claims 1, 4, and 12 of the 745 patent
on the ground of obviousness over Burton and Karedla, (iii) claim 2 on the
ground of obviousness over Karedla and Otterness, and (iv) claim 2 on the
ground of obviousness over Burton, Karedla, and Otterness.
FURTHER ORDERED that no other ground set forth in the Petition is
authorized for inter partes review; and
FURTHER ORDERED that pursuant to 35 U.S.C. 314(c) and
37 C.F.R. 42.4, notice is hereby given of the institution of a trial, which
commences on the entry date of this decision.
56
IPR2016-01643
Patent 6,775,745 B1
FOR PETITIONER:
P. Andrew Riley
James D. Stein
FINNEGAN, HENDERSON, FARABOW, GARRETT & DUNNER, LLP
andrew.riley@finnegan.com
james.stein@finnegan.com
IV745-IPR@finnegan.com
Jonathan Stroud
UNIFIED PATENTS INC.
jonathan@unifiedpatents.com
John R. King
Ted M. Cannon
KNOBBE, MARTENS, OLSON & BEAR, LLP
2jrk@knobbe.com
2tmc@knobbe.com
Tim R. Seeley
James R. Hietala
INTELLECTUAL VENTURES
tim@intven.com
jhietala@intven.com
57