Ex Parte Barsness et al

Patent Trial and Appeal Board

Mar 24, 2014

11928768 (P.T.A.B. Mar. 24, 2014)

UNITED STATES PATENT AND TRADEMARK OFFICE
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
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
11/928,768 10/30/2007 Eric Lawrence Barsness ROC920070247US1 7809
46797 7590 03/24/2014
IBM CORPORATION, INTELLECTUAL PROPERTY LAW
DEPT 917, BLDG. 006-1
3605 HIGHWAY 52 NORTH
ROCHESTER, MN 55901-7829
EXAMINER
SHANMUGASUNDARAM, KANNAN
ART UNIT PAPER NUMBER
2158
MAIL DATE DELIVERY MODE
03/24/2014 PAPER
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.
PTOL-90A (Rev. 04/07)

UNITED STATES PATENT AND TRADEMARK OFFICE
__________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
__________

Ex parte ERIC LAWRENCE BARSNESS, DAVID L. DARRINGTON,
AMANDA PETERS, and JOHN MATTHEW SANTOSUOSSO
__________

Appeal 2011-011708
Application 11/928,768
Technology Center 2100
__________

Before DONALD E. ADAMS, DEMETRA J. MILLS, and
JEFFREY N. FREDMAN, Administrative Patent Judges.

FREDMAN, Administrative Patent Judge.

DECISION ON APPEAL
This is an appeal1 under 35 U.S.C. § 134 involving claims to a method
of performing a garbage collection cycle on a parallel computing system
having a plurality of compute nodes. The Examiner rejected the claims as
anticipated and as obvious. We have jurisdiction under 35 U.S.C. § 6(b).
We reverse.

1 Appellants identify the Real Party in Interest as International
Business Machines Corporation (see App. Br. 2).
Appeal 2011-011708
Application 11/928,768

2
Statement of the Case
Background
“A garbage collector is software that runs concurrently with object-
oriented applications to free up memory as applications release objects”
(Spec. 1 ¶ 0002). The Specification teaches that “[k]eeping objects logically
separated allows the garbage collector to more aggressively check the
objects in the nursery since they are more likely to be short-lived, and only
periodically analyze the objects in the tenured space since they are more
likely to be long-lived” (Spec. 2 ¶ 0005).
The Claims
Claims 1-7, 9-15, and 17-23 are on appeal. Claim 1 is representative
and reads as follows:
1. A method of performing a garbage collection cycle on
a parallel computing system having a plurality of compute
nodes each running an instance of a computing job and a
garbage collector, comprising:
identifying, by the instance of the garbage collector
running on each of the plurality of compute nodes, an object
space in a memory of a respective compute node, wherein
the object space stores one or more objects allocated by the
instance of the computing job running on the respective
compute node;
evaluating each object in the object space to
determine whether the memory allocated to each respective
object is eligible to be collected;
upon determining that the memory allocated to a
given object is eligible to be collected, removing the given
object from the object space and returning memory allocated
to the given object to a pool;
determining a set of garbage collection statistics
associated with at least one evaluated object in the object
space; and
Appeal 2011-011708
Application 11/928,768

3
transmitting the set of garbage collection statistics to a
master garbage collector running on a compute node of the
parallel computing system, wherein the master garbage
collector distributes the transmitted set of garbage collection
statistics received from the garbage collector on each
respective compute node to the instance of the garbage
collector running on the other compute nodes, of the
plurality.

The issues
A. The Examiner rejected claims 1, 9, and 17 under 35 U.S.C. § 102(b)
as anticipated by Taura2 (Ans. 4-10).
B. The Examiner rejected claims 2-7, 10-15, and 18-23 under 35 U.S.C.
§ 103(a) as obvious over Taura and Agesen3 (Ans. 11-23).
Because both of these rejections turn on the same issues, we will
consider these rejections together.
The Examiner finds that Taura teaches all of the limitations of the
claims, including the limitation in claim 1 to
transmitting the set of garbage collection statistics to a
master garbage collector running on a compute node of the
parallel computing system, wherein the master garbage
collector distributes the transmitted set of garbage collection
statistics received from the garbage collector on each
respective compute node to the instance of the garbage
collector running on the other compute nodes, of the
plurality.

2 Kenjiro Taura & Akinori Yonezawa, An Effective Garbage
Collection Strategy for Parallel Programming Languages on
Large Scale Distributed-Memory Machines, PROCEEDINGS OF THE
SIXTH ACM SIGPLAN SYMPOSIUM ON PRINCIPLES AND PRACTICE
OF PARALLEL PROGRAMMING 264-275 (1997).
3 Agesen et al., US 2001/0044856 A1, published Nov. 22, 2001.
Appeal 2011-011708
Application 11/928,768

4
(Ans. 4-6.)
The Examiner finds that “heap size in Taura is a garbage collection
statistic because heap size is directly related to the garbage collection
process” (Ans. 27). The Examiner finds that “[s]ince heap size is increased
only when necessary to avoid too frequent garbage collections, then the heap
size is ‘derived from garbage collection activities performed by a garbage
collector on a processing node’ as required by the Appellant” (Ans. 28).
The Examiner finds that Taura teaches “wherein a set of garbage
collection statistics are determined (‘H’ is determined), transmitted, and then
distributed when in Taura, the processor i.e. node, has the largest local heap
size, i.e. M, and each processor is informed of that heap size i.e. during the
global collection” (Ans. 26-27).
The issue with respect to this rejection is: Does the evidence of record
support the Examiner’s conclusion that Taura anticipates claim 1?
Findings of Fact
1. The Specification teaches that:
The garbage collector 320 running on each node 112 may
collect specific information about the object types they are
tracking, such as percentage of objects of this type that are
long lived, average life time of an object, average life time
of short lived objects of this type, average life time of long
lived objects of this type, etc. The nodes then share this
information to the master node, which then distributes the
shared information back out to all of the nodes.

(Spec. 12 ¶ 0038.)
2. The Specification teaches that “the garbage collector may store
job creation statistics in the garbage collection statistics 530. Job creation
Appeal 2011-011708
Application 11/928,768

5
statistics may include, e.g., object type, date and time of object creation, and
name of method that creates the object” (Spec. 17 ¶ 0053).
3. Taura teaches “design and implementation of a garbage
collection scheme on large-scale distributed-memory computers” (Taura,
abstract).
4. Taura teaches
A global collection is invoked when a local collection
is not invoked by the criterion in the previous section and
when enough allocation has been done since the last global
collection. More precisely, letting H be the local heap size,
Al the amount of allocation done since the last local
collection, Ag the amount of allocation done since the last
global collection, and r a customizable parameter, we invoke
a global collection when:
Al ≤ H/r < Ag.
That is, when enough allocation has not been done
since the last local collection, but has been done since the
last global collection.

(Taura 269, col. 2.)
5. Taura teaches that
Each processor is informed of the largest local heap size
over all the processors. Let us call the value M, which is
made consistent at every global collection. When a
processor’s local heap size is smaller than cM where c is a
constant close to 1.0 (we currently set c to 0.8), the
processor expands the heap without invoking a garbage
collection.

(Taura 269, col. 1.)
Appeal 2011-011708
Application 11/928,768

6
6. Taura teaches that “if this policy would not be taken,
synchronized local collections would become unreasonably frequent” (Taura
269, col. 1).
Principles of Law
“In proceedings before the Patent and Trademark Office, the
Examiner bears the burden of establishing a prima face case of obviousness
based upon the prior art.” In re Fritch, 972 F.2d 1260, 1265 (Fed. Cir.
1992).
In addition, “obviousness requires a suggestion of all limitations in a
claim.” CFMT, Inc. v. Yieldup Int’l Corp., 349 F.3d 1333, 1342 (Fed. Cir.
2003).
Analysis
We begin with claim interpretation, since before a claim is properly
interpreted, its scope cannot be compared to the prior art. There are two
disputed phrases, “garbage collection statistic” and “wherein the master
garbage collector distributes the transmitted set of garbage collection
statistics received from the garbage collector on each respective compute
node to the instance of the garbage collector running on the other compute
nodes, of the plurality” in claim 1.
During prosecution, claim terms are given their broadest reasonable
interpretation as they would be understood by persons of ordinary skill in the
art in the light of the Specification. Therefore, we first turn to the
Specification to interpret these phrases.
Appeal 2011-011708
Application 11/928,768

7
“Garbage collection statistic”
The issue here is whether the broadest reasonable interpretation of
“garbage collection statistic” encompasses “heap size” as such a statistic.
Appellants contend that the “size of the heap on a given processing
node is not a garbage collection statistic” (App. Br. 13). Appellants contend
that while “Taura points out the heap size may be relevant to what type of
garbage collection cycle to perform, it is not a statistic derived from garbage
collection activities performed by a garbage collector on a given processing
node” (App. Br. 13).
The Specification teaches that the garbage collectors “collect specific
information about the object types they are tracking, such as percentage of
objects of this type that are long lived, average life time of an object” (Spec.
12 ¶ 0038; FF 1). The Specification also teaches that “the garbage collector
may store job creation statistics in the garbage collection statistics 530. Job
creation statistics may include, e.g., object type, date and time of object
creation, and name of method that creates the object” (Spec. 17 ¶ 0053; FF
2).
Thus, “garbage collection statistics” reasonably include a variety of
different statistics relating to garbage collection. When Taura teaches that
heap size is a statistic shared by processors (FF 5) which is obtained because
otherwise “synchronized local collections would become unreasonably
frequent” (FF 6) by the garbage collector, we agree with the Examiner that
“heap size” in Taura is reasonably interpreted as a “garbage collection
statistic.”
Appeal 2011-011708
Application 11/928,768

8
“wherein the master garbage collector distributes the transmitted set
of garbage collection statistics received from the garbage collector on each
respective compute node to the instance of the garbage collector running on
the other compute nodes, of the plurality”
The subtle issue for this phrase is whether it encompasses the situation
where a single node provides a “statistic” for distribution to the plurality of
other nodes, or whether the phrase requires that each of the plurality of
nodes provide a “statistic” for distribution to the plurality of nodes.
Appellants contend that “even if the heap size of a given processor is
presumed to be a ‘garbage collection statistic,’ nothing in Taura suggests
that the heap size of each node is transmitted to all the processing nodes, as
recited by the present claims” (App. Br. 14).
The Specification teaches that the “nodes then share this information
to the master node, which then distributes the shared information back out to
all of the nodes” (Spec. 12 ¶ 0038; FF 1). Thus, the Specification provides
descriptive support for the second alternative, where each node provides
“statistics” to the master node which are then distributed back to all of the
nodes (FF 1).
We find that claim 1 is reasonably interpreted as requiring that each
node provides information to the master node which redistributes to each of
the nodes, since the claim states that “wherein the master garbage collector
distributes the transmitted set of garbage collection statistics received from
the garbage collector on each respective compute node to the instance of the
garbage collector running on the other compute nodes.” This is most
reasonably understood as requiring that each node has a garbage collector
Appeal 2011-011708
Application 11/928,768

9
which has transmitted “statistics” to a master garbage collector. The claim
then requires that these are distributed back to the plurality of garbage
collection nodes.
Having interpreted claim 1, we agree with Appellants that while Taura
teaches distribution by a master node of a garbage statistic to each of the
other nodes (FF 5), Taura does not teach receiving garbage collection
statistics from each of the nodes. At best, Taura teaches receiving the heap
size statistic from a particular node. The Examiner has not established that
Taura teaches a master node which receives garbage collection statistics
from each of the respective nodes.
Conclusion of Law
The evidence of record does not support the Examiner’s conclusion
that Taura anticipates claim 1.
B. 35 U.S.C. § 103(a) over Taura and Agesen
This rejection relies upon the underlying anticipation rejection over
Taura. Having reversed the rejection of claim 1, we necessarily reverse this
obviousness rejection further including Agesen, since Agesen is not relied
upon to teach “wherein the master garbage collector distributes the
transmitted set of garbage collection statistics received from the garbage
collector on each respective compute node to the instance of the garbage
collector running on the other compute nodes, of the plurality” as required
by claim 1.

Appeal 2011-011708
Application 11/928,768

10
SUMMARY
In summary, we reverse the rejection of claims 1, 9, and 17 under
35 U.S.C. § 102(b) as anticipated by Taura.
We reverse the rejection of claims 2-7, 10-15, and 18-23 under
35 U.S.C. § 103(a) as obvious over Taura and Agesen.

REVERSED

cdc