Ex Parte Vasquez Lopez et al

Patent Trials and Appeals Board

Jan 24, 2019

13160963 - (D) (P.T.A.B. Jan. 24, 2019)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE
13/160,963 06/15/2011
69316 7590 01/28/2019
MICROSOFT CORPORATION
ONE MICROSOFT WAY
REDMOND, WA 98052
FIRST NAMED INVENTOR
Daniel Vasquez Lopez
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www .uspto.gov
ATTORNEY DOCKET NO. CONFIRMATION NO.
332826.01 4409
EXAMINER
MAHMOOD, REZWANUL
ART UNIT PAPER NUMBER
2164
NOTIFICATION DATE DELIVERY MODE
01/28/2019 ELECTRONIC
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
usdocket@microsoft.com
chriochs@microsoft.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte DANIEL VASQUEZ LOPEZ and WON YOO
Appeal2018-004232
Application 13/160,963
Technology Center 2100
Before JOHN A. JEFFERY, JOHNNY A. KUMAR, and
CARLL. SILVERMAN, Administrative Patent Judges.
JEFFERY, Administrative Patent Judge.
DECISION ON APPEAL
Appellants 1 appeal under 35 U.S.C. § 134(a) from the Examiner's
decision to reject claims 1-12 and 14--21. We have jurisdiction under
35 U.S.C. § 6(b). We affirm-in-part.
STATEMENT OF THE CASE
Appellants' invention evicts entries from a file handle cache. To this
end, entries in the file handle cache are analyzed and marked for eviction
while a shared lock is maintained on the cache. The entries are then
1 Appellants identify the real party in interest as Microsoft Technology
Licensing, LLC. App. Br. 1.
Appeal2018-004232
Application 13/160,963
removed while an exclusive lock is maintained on the cache. See generally
Abstract. Claims 1, 12, and 16 are illustrative:
1. A method for evicting entries from a file handle cache, each
entry in the file handle cache storing a file handle associated with an open
file in a file system, the method comprising:
analyzing each entry in the file handle cache to identify entries to be
marked for eviction and marking the identified entries for eviction while
maintaining a shared lock on the entire file handle cache, the shared lock
allowing the file handle cache to be concurrently accessed by one or more
processes to obtain file handles for use in servicing requests for content from
the file system; and
removing the entries marked for eviction from the file handle cache
while maintaining an exclusive lock on the entire file handle cache, the
exclusive lock preventing the file handle cache from being concurrently
accessed by the one or more processes.
12. A system, comprising:
at least one processor;
memory connected to the at least one processor;
a file system stored at least in part in the memory and executed by the
at least one processor; and
a content serving system stored at least in part in the memory and
executed by the at least one processor, the content serving system being
configured to service requests for content by accessing files on the file
system, the content serving system including:
a file handle cache having a plurality of entries, each entry in the file
handle cache storing a file handle associated with an open file in the file
system, the file handle cache being accessible to obtain file handles for
accessing open files on the file system; and
file handle cache management logic that is configured to analyze a
time to live counter assigned to each entry in the file handle cache and to
mark an entry for eviction if the time to live counter assigned thereto has
expired, the value of the time to live counter that is assigned to each entry
in the file handle cache being adaptively determined based on a measure
of utilization of a system memory that stores contents of each open file
associated with each file handle stored in the file handle cache, the system
memory being a different entity than the file handle cache.
2
Appeal2018-004232
Application 13/160,963
16. A system, comprising:
at least one processor;
memory connected to the at least one processor;
a file system stored at least in part in the memory and executed by the
at least one processor; and
a content serving system stored at least in part in the memory and
executed by the at least one processor, the content serving system being
configured to service requests for content by accessing files on the file
system, the content serving system including:
a file handle cache having a plurality of entries, each entry in the file
handle cache storing a file handle associated with an open file in the file
system, the file handle cache being accessible to obtain file handles for
accessing open files on the file system; and
file handle cache management logic that is configured to analyze a hit
count associated with each entry in the file handle cache and to mark an
entry for eviction if the hit count is less than a statistical threshold, the hit
count associated with each entry representing a number of times that the
entry was accessed to service a request for content from the file system, the
hit count associated with each entry being compared to the same statistical
threshold, the statistical threshold being determined based on a combination
of the hit counts associated with all of the entries in the file handle cache.
THE REJECTIONS
The Examiner rejected claim 1 under 35 U.S.C. § 103 as unpatentable
over Wintergerst (US 2006/0064549 Al; published Mar. 23, 2006), Pawar
(US 8,204,871 Bl; issued June 19, 2012), and Ezra (US 7,143,393 Bl;
issued Nov. 28, 2006). Final Act. 13-15. 2
2 Throughout this opinion, we refer to (1) the Final Rejection mailed June
22, 2017 ("Final Act."); (2) the Appeal Brief filed August 31, 2017 ("App.
Br."); (3) the Examiner's Answer mailed January 11, 2018 ("Ans."); and (4)
the Reply Brief filed March 12, 2018 ("Reply Br.").
3
Appeal2018-004232
Application 13/160,963
The Examiner rejected claims 2 and 3 under 35 U.S.C. § 103 as
unpatentable over Wintergerst, Pawar, Ezra, and Coulson (US 2007 /023394 7
Al; published Oct.4.2007). Final Act. 16-17.
The Examiner rejected claims 4--11 under 35 U.S.C. § 103 as
unpatentable over Wintergerst, Pawar, Ezra, and Petev (US 2006/0248285
Al; published Nov. 2, 2006). Final Act. 18-22.
The Examiner rejected claims 12, 14, 15, and 21 under 35 U.S.C.
§ 103 as unpatentable over Wintergerst, Pa war, Petev, and Steely (US
2005/0154805 Al; published July 14, 2005). Final Act. 22-27.
The Examiner rejected claims 16-20 under 35 U.S.C. § 103 as
unpatentable over Wintergerst, Pawar, Petev, and Matthews (US
2007 /0005889 Al; published Jan. 4, 2007). Final Act. 27-32.
THE REJECTION OVER WINTERGERST, PA WAR, AND EZRA
Regarding independent claim 1, the Examiner finds that Wintergerst
discloses, among other things, a method for evicting entries from a cache,
where each cache entry is analyzed to identify entries to be marked for
eviction, and so marking the identified entries. Final Act. 13-14. Although
the Examiner acknowledges that Wintergerst lacks ( 1) a file handle cache
storing a file handle associated with an open file, and (2) shared and
exclusive locks on the file handle cache, the Examiner cites Pawar for
teaching these features. Id. 14--15. The Examiner also acknowledges that
the Wintergerst/Pawar system does not lock an entire cache, but cites Ezra to
cure that deficiency. Id. 15. Based on these collective teachings, the
Examiner concludes that the claim would have been obvious. Id.
4
Appeal2018-004232
Application 13/160,963
Appellants argue that the cited prior art does not teach or suggest
analyzing each entry in a file handle cache to identify entries to be marked,
and marking the entries for eviction while maintaining a shared lock on the
entire file handle cache, as claimed. App. Br. 8-13; Reply Br. 3-5. 3
According to Appellants, not only does the Examiner improperly dissect
the disputed limitation and address various portions of this limitation "in a
vacuum," the Examiner's reliance on the cited prior art references is said
to be misplaced in any event. See App. Br. 8-11. Although Appellants
acknowledge that Wintergerst's system evicts cache entries that are so
marked, Appellants emphasize that Wintergerst does mark these entries
while a shared lock is maintained on any portion of a cache, let alone an
entire cache as claimed. App. Br. 9. Appellants add that although Pawar's
locks are applied to certain blocks of a specified file stored in a cache, and
Ezra maintains an exclusive lock on an entire cache, the references do not
maintain a shared lock on an entire file handle cache, let alone mark cache
entries for deletion while maintaining the shared lock, as claimed. App. Br.
10-11. Lastly, Appellants argue that there is no apparent reason to combine
the references as the Examiner proposes apart from improper hindsight, and
that the Examiner merely provides conclusory statements that it would have
been obvious to combine the teachings of the cited references. App. Br. 12-
13; Reply Br. 4--5.
3 Although the Reply Brief is unpaginated (unlike the Appeal Brief), we
nonetheless refer to the Reply Brief' s pages in the order that they appear in
the record.
5
Appeal2018-004232
Application 13/160,963
ISSUE
Under§ 103, has the Examiner erred in rejecting claim 1 by finding
that Wintergerst, Pawar, and Ezra collectively would have taught or
suggested analyzing each entry in a file handle cache to identify entries to be
marked for eviction, and so marking the entries while maintaining a shared
lock on the entire file handle cache?
ANALYSIS
We begin by noting a key temporal aspect of claim 1, namely that the
identified entries are marked for eviction while maintaining a shared lock on
the entire file handle cache. Our emphasis on the term "while" underscores
that this limitation specifies when entries are marked as Appellants indicate
(App. Br. 9), namely simultaneously with maintaining the shared lock.
Given this temporal requirement, the Examiner's findings in this
regard are problematic on this record. In the rejection, the Examiner finds
that Wintergerst' s system ( 1) analyzes each entry in a cache to be marked for
eviction, and (2) marks the identified entries for eviction "while maintaining
a shared ... on the ... cache .... " Final Act. 13 ( emphasis added). These
omissions indicated by the ellipses are telling, for it is unclear from these
findings what exactly Wintergerst's system maintains while marking occurs.
On page 5 of the Answer, the Examiner cites Wintergerst's
paragraphs 12, 13, 15, 68, 75, and 101, and equates Wintergerst's (1)
marking entries in a shared cache for eviction, and (2) maintaining a shared
cache with marking identified entries for eviction while maintaining shared
access on the cache. Despite the claim reciting that marking occurs while
maintaining a shared lock, the Examiner nevertheless cites Wintergerst's
6
Appeal2018-004232
Application 13/160,963
paragraphs 16, 65, 68, 75, and 81, and finds that Wintergerst's shared cache
allows read-only access to cache entries or locks them out entirely. See Ans.
6. Although Wintergerst' s paragraph 81 describes a hash map that can
indicate whether an item is locked, we cannot say that marking identified
entries while maintaining a shared lock on an entire file handle cache would
have been obvious from the references' cited passages.
The Examiner's reliance on the other references fares no better in
this regard. Although Pawar teaches placing shared range locks on a file
specified by a file handle in column 8, lines 22 to 57 and step 131 of Figure
5, and Ezra teaches locking an entire cache data structure in column 21, lines
47 to 58, the cited prior art collectively still does not teach or suggest
marking cache entries for eviction while maintaining a shared lock on an
entire file handle cache, as claimed.
So even if the cited references were combinable as the Examiner
proposes, they still do not teach or suggest the disputed temporal limitation.
Accordingly, on the record before us, we are persuaded that the Examiner
erred in rejecting claim 1.
THE REJECTIONS OF CLAIMS 2-11
Because the Examiner has not shown that the additional cited prior art
cures the foregoing deficiencies regarding the rejection of independent claim
1, we do not sustain the obviousness rejections of dependent claims 2-11
(Final Act. 16-22) for similar reasons.
7
Appeal2018-004232
Application 13/160,963
THE REJECTION OVER WINTERGERST, PAW AR, PETEY, AND
STEELY
We also do not sustain the Examiner's obviousness rejection of
independent claim 12 over Wintergerst, Pawar, Petev, and Steely. Final Act.
22-27. Claim 12 recites, in pertinent part, file handle cache management
logic configured to analyze a time-to-live4 counter assigned to each file
handle cache entry and to mark an entry for eviction if the assigned counter
has expired, where the assigned counter's value is adaptively determined
based on a measure of utilization of a system memory that stores contents of
each open file associated with each file handle stored in the file handle
cache.
Our emphasis underscores a key aspect of the recited counter, namely
that its value is adaptively determined based on a particular measure of
system memory utilization, where the system memory differs from the file
handle cache. As Appellants explain, by adapting time-to-live values based
on system memory utilization, file handles can be evicted from the file
handle cache faster when system memory utilization is high, thus causing
files to be closed and reducing system memory utilization. App. Br. 20-21.
In the rejection, the Examiner finds that Wintergerst's cache
management logic "is configured to analyze ... each entry in the ... cache
and to mark an entry for eviction ... adaptively determined based on a
measure of utilization of a system memory." Final Act. 23-24 ( citing
Wintergerst ,r,r 12, 13, 15; emphasis added). Our emphasis underscores that
the Examiner relies on Wintergerst for the recited adaptive determination,
4 Although the term "time to live" is not hyphenated in the claim, we
nonetheless hyphenate it for clarity and grammatical form.
8
Appeal2018-004232
Application 13/160,963
yet acknowledges that Wintergerst lacks a time-to-live counter. See Final
Act. 23-25. That is, the Examiner apparently finds that Wintergerst
adaptively determines marking cache entries for eviction-not adaptively
determining a counter value as Appellants indicate. App. Br. 16.
To be sure, Petev's paragraphs 95 and 124 indicate that a time-to-live
counter value can be associated with a cache entry, and that an object can be
evicted if it resides in cache longer than the time-to-live. But Petev says
nothing about adaptively determining this value, let alone basing this
adaptive determination on a particular measure of system memory
utilization, where the system memory differs from the file handle cache as
claimed. Accord App. Br. 18 (noting this deficiency). Nor do Pawar and
Steely cure this deficiency. Although Pawar's extent cache can include file
handles as suggested in column 8, lines 22 to 28, and Steely's paragraph 2
notes that processors can have their own memory cache separate from
main system memory, these general teachings do not cure the foregoing
deficiencies regarding adaptively determining the recited counter value as
claimed. To the extent that the Examiner finds that the cited references
collectively teach the recited counter value adaptive determination, we
disagree.
So even if the cited references were combinable as the Examiner
proposes, they still do not teach or suggest the recited counter value adaptive
determination. Accordingly, we are persuaded that the Examiner erred in
rejecting independent claim 12, and dependent claims 14, 15, and 21 for
similar reasons.
9
Appeal2018-004232
Application 13/160,963
THE REJECTION OVER WINTERGERST, PAW AR, PETEY, AND
MATTHEWS
We sustain the Examiner's obviousness rejection of independent
claim 16 over Wintergerst, Pawar, Petev, and Matthews. Final Act. 27-32.
Claim 16 recites, in pertinent part, file handle cache management logic
configured to (1) analyze a hit count associated with each cache entry,
and (2) mark an entry for eviction if the hit count is less than a statistical
threshold, where the hit count represents a number of times the entry was
accessed, and the threshold is determined based on a combination of hit
counts associated with all entries in the cache.
The Examiner cites Petev's paragraphs 113 to 115 and 119 to 123 for
teaching much of the recited cache management logic functionality,
including (1) analyzing hit counts associated with each cache entry, and (2)
marking an entry for eviction if the hit count is less than a statistical
threshold. Final Act. 29; Ans. 17-18. We see no error in these findings.
Petev uses a counter to measure the number of times an object is accessed,
and based on this measurement, places the object in a particular location in a
queue. See Petev ,r,r 113-115. Notably, the object with the lowest counter
value (i.e., the object that is the least frequently used) is at the bottom of
the queue and, consequently, will be evicted next from the cache, thus
effectively marking that entry for eviction. See id.
Although the Examiner acknowledges that Petev' s threshold is not
determined based on a combination of hit counts associated with all entries
in the cache, we see no error in the Examiner's reliance on Matthews for at
least suggesting such a determination. See Final Act. 30-31; Ans. 18-19
(citing Matthews ,r,r 14, 19, 23, 26-28). To be sure, Matthews' system uses
10
Appeal2018-004232
Application 13/160,963
the threshold length of data to determine which data requests are cached.
See Matthews ,r 19. But a key aspect of this determination is that the
threshold length can be changed dynamically based on hit rates for all data
in the cache as well as data rejected from being read from or written to the
cache. See Matthews ,r,r 19, 26-27.
Given Matthews' fundamental teaching of changing a threshold based
on a combination of hit counts for all cached entries, we see no reason why
such a determination could not be applied to a hit count threshold, such as
that in Petev, as the Examiner proposes to, among other things, manage the
cache efficiently by setting a threshold dynamically based on the total hit
counts. See Final Act. 30-31; Ans. 18-20. That Matthews pertains to a
cache entry scheme unlike Petev' s cache eviction scheme as Appellants
indicate (App. Br. 25) is of no consequence here, for both schemes
ultimately determine whether data will be stored in a cache based at least
partly on hit count. Therefore, the Examiner's proposed combination is
supported by articulated reasoning with some rational underpinning to
justify the Examiner's obviousness conclusion.
We reach the same conclusion regarding the Examiner's reliance on
Wintergerst and Pawar for at least suggesting the respective features for
which whey were cited, as well as the Examiner's articulated reason to
combine the cited references. Final Act. 27-31; Ans. 15-20. Despite
Appellants' arguments to the contrary (App. Br. 26-27; Reply Br. 8-9), the
Examiner's articulated reason to combine the references is not merely
conclusory as Appellants assert, but rather has at least some rational
underpinning to justify the Examiner's obviousness conclusion.
11
Appeal2018-004232
Application 13/160,963
Therefore, we are not persuaded that the Examiner erred in rejecting
claim 16, and claims 17-20 not argued separately with particularity.
CONCLUSION
Under§ 103, the Examiner erred in rejecting claims 1-12, 14, 15, and
21, but did not err in rejecting claims 16-20.
DECISION
We affirm-in-part the Examiner's decision to reject claims 1-12 and
14--21.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(l)(iv).
AFFIRMED-IN-PART
12