Ex Parte Yoder

Board of Patent Appeals and Interferences

Feb 22, 2010

10832758 (B.P.A.I. Feb. 22, 2010)

UNITED STATES PATENT AND TRADEMARK OFFICE
__________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
__________

Ex parte MICHAEL E. YODER
__________

Appeal 2009-005866
Application 10/832,758
Technology Center 2100
__________

Decided: February 22, 2010
__________

Before JAY P. LUCAS, THU A. DANG, and STEPHEN C. SIU,
Administrative Patent Judges.

SIU, Administrative Patent Judge.

DECISION ON APPEAL
STATEMENT OF THE CASE
This is a decision on appeal under 35 U.S.C. § 134(a) from the
Examiner’s rejection of claims 1, 6-14, 19-27, and 30-36. Claims 2-5, 15-
18, 28, and 29 are canceled. We have jurisdiction under 35 U.S.C. § 6(b).
We affirm.
Appeal 2009-005866
Application 10/832,758

2
Invention
The invention relates to “various approaches for allocating memory
objects in a non-uniform memory access (NUMA) system” (Spec. 3,
¶ [0006]).
Independent claim 1 is illustrative:
1. A processor-implemented method for allocating
memory objects in a non-uniform memory access
(NUMA) system, wherein the NUMA system includes a
plurality of nodes, and each node includes at least one
processor coupled to a memory subsystem via a local
bus, and the NUMA system includes an interleave
memory, comprising:
establishing at least one instance of a data structure of a
first type including a plurality of locality definitions,
wherein each instance of the first type data structure has
an associated set of program-configurable attributes used
in controlling allocation of memory objects via the
instance of the first type data structure, and each locality
definition being selectable via a locality identifier and
designating a memory subsystem in the NUMA system;
in response to a request from a processor in the NUMA
system for allocation of memory objects via the instance
of the first type data structure and specifying a locality
identifier, allocating to the processor memory objects
from the memory subsystem designated by the locality
definition as referenced by the locality identifier;
in response to a request from a processor for allocation of
memory objects via an instance of the first type data
structure and not specifying a locality identifier,
allocating to the processor memory objects from the
Appeal 2009-005866
Application 10/832,758

3
memory subsystem that is in the node of the requesting
processor;
establishing at least a first instance of a data structure of a
second type including a single locality definition,
wherein each instance of the second type data structure
has an associated set of program-configurable attributes
used in controlling allocation of memory objects via the
first instance of the second type data structure, and the
single locality definition of the first instance references a
memory subsystem in a node;
in response to a request from a processor in the NUMA
system for allocation of memory objects via the first
instance of the second type data structure, allocating to
the processor memory objects from the memory
subsystem referenced by the single locality definition
consistent with attributes of the first instance of the data
structure of a second type;
maintaining in each instance of the first type and second
type data structures, respective lists of free memory
objects for each node in the NUMA system for allocation
requests;
establishing at least a second instance of the data
structure of the second type, wherein the single locality
definition of the second instance references the interleave
memory in the NUMA system; and
in response to a request from a processor in the NUMA
system for allocation of memory objects via the second
instance, allocating to the processor memory objects
from interleave memory consistent with attributes of the
second instance of the data structure of the second type.
Appeal 2009-005866
Application 10/832,758

4
References
The Examiner relies upon the following references as evidence in
support of the rejections:
Fung US 4,924,375 May 8, 1990
Stevens US 6,336,177 B1 Jan. 1, 2002
Roger L. Adema and Carla S. Ellis, Memory Allocation Constructs to
Complement NUMA Memory Management, Proceedings of the Third IEEE
Symposium on Parallel and Distributed Processing, pp. 878-885 (Dec. 1991)
(“Adema”).
Rejections
Claims 1, 6-14, 19-27, and 30-36 are rejected under 35 U.S.C.
§ 103(a) as being unpatentable over Stevens, Adema, and Fung.
ISSUE 1
The Examiner finds that “Stevens discloses selecting a locality
definition that specifies a specific I/O [input/output] device to have a strong
affinity to [while] Fung teaches that interleave memory is useful in replacing
general memory” (Ans. 10).
Appellant submits that Stevens and Fung would not have taught or
suggested “interleave memory [that can] be one locality definition in
combination with the claimed single and multiple locality definitions” (App.
Br. 7-8).
Issue: Did Appellant demonstrate that the Examiner erred in
concluding that Stevens and Fung would have taught or suggested use of
Appeal 2009-005866
Application 10/832,758

5
interleave memory as a locality definition when allocating memory in a
NUMA system?
ISSUE 2
The Examiner interprets “the phrasing ‘maintaining in each instance
of the . . . data structures’ as requiring the instances of the data structures
maintain or somehow update / manage the free lists” (Ans. 12).
Appellant argues that “for each of Adema’s free lists, the memory that
is referenced is the memory available on that respective node” (App. Br. 10).
Issue: Did Appellant demonstrate that the Examiner erred in
concluding that Stevens, Adema, and Fung would have taught or suggested
maintaining in data structures lists of free memory objects for nodes in a
NUMA system?
ISSUE 3
The Examiner finds that “Adema discloses that the USAlloc system
will try to allocate memory on the node on which it is invoked, and if that
fails . . . then it attempts to allocate memory from nearby nodes” (Ans. 14-
15).
Appellant argues that there is “clearly no ‘effective’ adding of
memory blocks from one free list to another followed by removal” (App.
Br. 12).
Issue: Did Appellant demonstrate that the Examiner erred in
concluding that Stevens and Adema would have taught or suggested adding
memory objects from a designated memory subsystem to a free list in
Appeal 2009-005866
Application 10/832,758

6
response to the free list having an insufficient number of memory objects to
satisfy a first request?
ISSUE 4
The Examiner finds that “Adema discloses a request being able to be
completed when it does not specify a locality parameter, and that the request
will just attempt to allocate memory from the instant node” (Ans. 15). The
Examiner also finds that if the node “had an interleave memory, then the
request to allocate memory that didn’t have a locality parameter would
potentially receive memory allocated from an interleave memory” (Ans. 16).
Appellant submits that Adema’s “UsAlloc function is for allocating
memory, not for establishing the data structure used in controlling allocation
of memory” (App. Br. 13). Appellant further argues that “there is no
suggestion that if no locality is specified in a request, the response is to
establish a locality definition that references interleave memory” (id.)
Issue: Did Appellant demonstrate that the Examiner erred in
concluding that Stevens, Adema, and Fung would have taught or suggested
creating an instance of a data structure for controlling allocation of memory
objects, the data structure including a single locality definition that
references interleave memory, in response to a creation request that does not
specify a locality identifier?
FINDINGS OF FACT
The following Findings of Fact (FF) are shown by a preponderance of
the evidence.
Appeal 2009-005866
Application 10/832,758

7
1. Stevens teaches that “by specifying MLDs [memory
locality domains] and MLDSETs [MLD sets], rather than
physical memory nodes, application programs can be executed
on different computer systems regardless of the particular node
configuration and physical node topology employed by the
system” (col. 2, ll. 45-49).
2. Stevens teaches an MLD 802 with a radius of 0 (fig 8),
indicating “that no network hops are required to access memory
from a thread attached to its center node” (col. 9, ll. 25-27).
3. Fung teaches interleaving sequential pages of memory
“between memory banks so that memory accesses which are a
page apart will be to two different memory banks” (abstract).
4. Adema teaches a “Uniform System [that] maintains a
singly-linked list of blocks of available memory (a ‘free list’)
for each node in the cluster” (p. 880).
5. Adema teaches that “[w]hen UsAlloc is called, the
Uniform System first attempts to allocate memory on the node
on which UsAlloc is being invoked. If unsuccessful, UsAlloc
then tries to allocate the memory elsewhere by examining the
free list of each node in the cluster” (p. 880).
Appeal 2009-005866
Application 10/832,758

8
PRINCIPLES OF LAW
Claim interpretation
“In the patentability context, claims are to be given their broadest
reasonable interpretations. . . . [L]imitations are not to be read into the
claims from the specification.” In re Van Geuns, 988 F.2d 1181, 1184 (Fed.
Cir. 1993) (citations omitted). A claim meaning is reasonable if one of
ordinary skill in the art would understand the claim, read in light of the
specification, to encompass the meaning. See In re American Academy of
Science Tech Center, 367 F.3d 1359, 1364 (Fed. Cir. 2004). Any special
meaning assigned to a term “must be sufficiently clear in the specification
that any departure from common usage would be so understood by a person
of experience in the field of the invention.” Multiform Desiccants Inc. v.
Medzam Ltd., 133 F.3d 1473, 1477 (Fed. Cir. 1998).
Obviousness
The question of obviousness is resolved on the basis of underlying
factual determinations including (1) the scope and content of the prior art,
(2) any differences between the claimed subject matter and the prior art, and
(3) the level of skill in the art. Graham v. John Deere Co., 383 U.S. 1, 17-
18 (1966). One cannot show nonobviousness by attacking references
individually where the rejections are based on combinations of references.
In re Merck & Co., Inc., 800 F.2d 1091, 1097 (Fed. Cir. 1986).
Appeal 2009-005866
Application 10/832,758

9
ANALYSIS
Issue 1
Appellant disputes the Examiner’s conclusion that Stevens and Fung
would have taught or suggested the claimed use of interleave memory as a
locality definition. We find Appellant’s arguments unpersuasive.
Stevens teaches the use of MLDs by application programs, abstracting
the particular node configurations and physical node topologies of different
system (FF 1). An MLD can have a radius of 0, indicating that the memory
to use is local to a node (FF 2). Fung teaches interleaving of memory
between memory banks (FF 3).
Appellant suggests the use of interleave memory as an MLD would
not have been obvious to an artisan because Fung described interleave
memory well before the time of the Stevens disclosure (App. Br. 10).
However, Appellant offers no evidence, other than chronology, to show that
an artisan would think it necessary to detail the types of memory that could
be used with Stevens’ teachings. Furthermore, we agree with the
Examiner’s undisputed finding that “Stevens is primarily concerned with the
overall framework of the multiprocessing system” (Ans. 12). Therefore, we
conclude that Stevens and Fung would have taught or suggested use of
interleave memory (as taught by Fung) as a locality definition (interleave
memory on a node referenced by an MLD with radius 0) when allocating
memory in a NUMA system (as taught by Stevens).
For at least these reasons, we conclude that Appellant has not
sustained the requisite burden on appeal in providing arguments or evidence
Appeal 2009-005866
Application 10/832,758

10
persuasive of error in the Examiner’s 35 U.S.C. § 103(a) rejection of claims
1, 6-14, and 19-27.
Issue 2
Appellant argues that the Examiner erred in concluding that Stevens,
Adema, and Fung would have taught or suggested maintaining in data
structures lists of free memory objects for nodes in a NUMA system. We
find Appellant’s arguments unpersuasive.
Appellant submits “that the claim language does indicate that the free
lists are ‘in’ the data structures” (Reply Br. 3). However, the Examiner’s
interpretation of maintaining—somehow updating or managing—is
reasonably broad (Ans. 12).
Stevens teaches allocation of memory using MLDs and MLDSETs
(FF 1) while Adema teaches maintenance of free lists for nodes (FF 4).
Using an MLD or MLDSET to allocate memory would affect the free
memory on nodes identified by the MLD or MLDSET. Therefore, Stevens
and Adema would have taught or suggested maintaining (affecting through
use of) in data structures (MLDs and MLDSETs) lists of free memory
objects for nodes in a NUMA system (lists of free memory on identified
nodes).
For at least these reasons, we conclude that Appellant has not
sustained the requisite burden on appeal in providing arguments or evidence
persuasive of error in the Examiner’s 35 U.S.C. § 103(a) rejection of claims
1, 6-14, 19-27, and 30-36.
Appeal 2009-005866
Application 10/832,758

11
Issue 3
Appellant and the Examiner disagree on the implications of Adema’s
USAlloc system. We agree with the Examiner’s findings and conclusion.
When a memory cannot be allocated on a node in Adema, the
Uniform System attempts to allocate memory from other nodes (FF 5).
Furthermore, Adema maintains a list of blocks of available memory (FF 4).
This would have taught or suggested to an artisan the steps of maintaining a
free list, including obtaining memory resources when necessary. Therefore,
Adema would have taught or suggested adding memory objects (allocating
memory) from a designated memory subsystem (a node) to a free list
(maintaining a free list) in response to the free list having an insufficient
number of memory objects to satisfy a first request (allocating in response to
in sufficient memory resources).
For at least these reasons, we conclude that Appellant has not
sustained the requisite burden on appeal in providing arguments or evidence
persuasive of error in the Examiner’s 35 U.S.C. § 103(a) rejection of claims
8-11, 21-24, and 34-36.
Issue 4
Appellant challenges the Examiner’s conclusion that Stevens, Adema,
and Fung would have taught or suggested use of interleave memory as a
default locality identifier. We find Appellant’s arguments unpersuasive.
Stevens teaches data structures (MLDs) to manage memory allocation
in a NUMA system (FF 1). An MLD can have a radius of 0, and thus point
to memory on a single node (FF 2). Adema teaches allocating from memory
Appeal 2009-005866
Application 10/832,758

12
local to a memory allocation requestor by default (FF 5). Fung teaches
interleave memory (FF 3). These references would have taught or suggested
creating an instance of a data structure for controlling allocation of memory
objects (MLDs), the data structure including a single locality definition (with
a radius of 0) that references interleave memory (the node having interleave
memory), in response to a creation request that does not specify a locality
identifier (i.e., a default locality as taught by Adema).
For at least these reasons, we conclude that Appellant has not
sustained the requisite burden on appeal in providing arguments or evidence
persuasive of error in the Examiner’s 35 U.S.C. § 103(a) rejection of claims
12, 25, and 31.
CONCLUSIONS OF LAW
Based on the findings of facts and analysis above, we conclude that
Appellant has not demonstrated:
1. that the Examiner erred in concluding that Stevens and Fung would
have taught or suggested use of interleave memory as a locality
definition when allocating memory in a NUMA system (Issue 1);
2. that the Examiner erred in concluding that Stevens, Adema, and
Fung would have taught or suggested maintaining in data
structures lists of free memory objects for nodes in a NUMA
system (Issue 2);
3. that the Examiner erred in concluding that Stevens and Adema
would have taught or suggested adding memory objects from a
Appeal 2009-005866
Application 10/832,758

13
designated memory subsystem to a free list in response to the free
list having an insufficient number of memory objects to satisfy a
first request (Issue 3); and
4. that the Examiner erred in concluding that Stevens, Adema, and
Fung would have taught or suggested creating an instance of a data
structure for controlling allocation of memory objects, the data
structure including a single locality definition that references
interleave memory, in response to a creation request that does not
specify a locality identifier (Issue 4).
DECISION
We affirm the Examiner’s decision rejecting claims 1, 6-14, 19-27,
and 30-36 under 35 U.S.C. § 103(a).
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED

msc
HEWLETT-PACKARD COMPANY
Intellectual Property Administration
3404 E. Harmony Road
Mail Stop 35
FORT COLLINS CO 80528