Ex Parte Branson et al

Patent Trial and Appeal Board

Oct 26, 2016

12821784 (P.T.A.B. Oct. 26, 2016)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR
12/821,784 06/23/2010 Michael J. Branson
46296 7590 10/28/2016
MARTIN & ASSOCIATES, LLC
P.O. BOX548
CARTHAGE, MO 64836-0548
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.
ROC920100043US1 4549
EXAMINER
DAO,TUANC.
ART UNIT PAPER NUMBER
2193
NOTIFICATION DATE DELIVERY MODE
10/28/2016 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):
derekm@ideaprotect.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte MICHAEL J. BRANSON
and JOHN M. SANTOSUOSSO
Appeal2015-000104
Application 12/821, 7 84
Technology Center 2100
Before ALLEN R. MacDONALD, JOHN F. HORVATH, and
KEVIN C. TROCK, Administrative Patent Judges.
HORVATH, Administrative Patent Judge.
DECISION ON APPEAL
STATEMENT OF THE CASE
Appellants seek review, under 35 U.S.C. § 134(a), of the Examiner's
rejection of claims 1, 2, 4--8, 16, 17, and 19-22. We have jurisdiction under
35 U.S.C. § 6(b).
We AFFIRM.
Appeal2015-000104
Application 12/821,784
SUMMARY OF THE INVENTION
The invention is directed to dynamically allocating a job on a multi-
nodal, parallel computer system. Spec. i-f 2.
Claim 1, reproduced below, is illustrative of the claimed subject
matter:
1. An apparatus comprising:
a plurality of nodes of a multi-nodal computer system,
wherein the plurality of nodes are connected by a plurality of
networks, where each of the plurality of nodes has at least one
central processing unit (CPU) coupled to a memory;
an application having a plurality of jobs, each with at least
one processing unit executing on the plurality of nodes;
a collector collecting metrics of the system, nodes,
application, jobs and processing units in order to determine how
to allocate the jobs on the system;
a job optimizer dynamically changing the allocation of
processing units on the plurality of nodes based on the collected
metrics, wherein the job optimizer dynamically changes the
allocation of the processing units by splitting a job running on a
first node into multiple jobs allocated to different nodes of the
plurality of nodes, wherein an interprocess communication
between processing units of the job is within a node before
splitting the job, and after splitting the job the interprocess
communication is then handled by the multi-nodal computer
system over a communication link between the different nodes.
F ellenstein
Dillenberger
Kaminsky
Abu-Ghazaleh
Barsness
Votta
REFERENCES
US 2006/0152756 Al
US 2008/0216087 A 1
US 2009/0048998 Al
US 2009/0083390 Al
US 2009/0313636 Al
US 2010/0011254 Al
2
July 13, 2006
Sept. 4, 2008
Feb. 19,2009
Mar. 26, 2009
Dec. 17, 2009
Jan. 14, 2010
Appeal2015-000104
Application 12/821,784
REJECTIONS
Claims 1, 16, and 21 stand rejected under 35 U.S.C. § 103(a) as
unpatentable over Abu-Ghazaleh and Fellenstein. Final Act. 2.
Claims 2 and 17 stand rejected under 35 U.S.C. § 103(a) as
unpatentable over Abu-Ghazaleh, Fellenstein, and Barsness. Final Act. 9.
Claim 4 stands rejected under 35 U.S.C. § 103(a) as unpatentable over
Abu-Ghazaleh, Fellenstein, and Dillenberger. Final Act. 10.
Claims 5, 6, 8, 19, 20, and 22 stand rejected under 35 U.S.C. § 103(a)
as unpatentable over Abu-Ghazaleh, Fellenstein, and Votta. Final Act. 11.
Claim 7 stands rejected under 35 U.S.C. § 103(a) as unpatentable over
Abu-Ghazaleh, Fellenstein, Votta, and Kaminsky. Final Act. 15.
ANALYSIS
Claims 1, 2, 4, 8, 16, 17, 21, and 22
Issue 1: Whether Abu-Ghazaleh teaches or suggests a job
optimizer that dynamically changes the allocation of processing
units by splitting a job running on a first node into multiple jobs
allocated to different nodes as required by claim 1.
The Examiner finds Abu-Ghazaleh (hereafter "Abu") teaches or
suggests this limitation. Final Act. (citing, inter alia, Abu i-fi-f 17, 125); Ans.
4--9 (citing, inter alia, Abu i-fi-f 17, 39, 94, 125). Appellants argue that
although Abu teaches partitioning a job into sub-tasks and assigning
different sub-tasks to different nodes, "Abu does not teach or suggest
splitting a job running on a first node into multiple jobs allocated to different
nodes." App. Br. 7. That is, Appellants argue that the allocation of tasks
3
Appeal2015-000104
Application 12/821,784
and subtasks in Abu is predetermined and fixed, and "[t]here is nothing
about splitting a job as claimed." Id. We are not persuaded by Appellants'
arguments.
Abu discloses a self-organizing, distributed computing grid that can
"tolerate a significant portion of its participating resources to be dynamically
added or removed, even those that are being used by active computations."
Abu i-f 17 (emphasis added). Abu divides jobs into tasks, and allocates tasks
to the nodes of the grid in order to "accomplish [the] tasks which are
partitioned to [the] various nodes, and also subject to maintenance in the
event that the underlying resources change in availability or use." Id.
(emphasis added). Abu teaches that "a job is a set of tasks ... carried out on
a set of resources, and a task is an atomic operation to be performed on a
resource." Id. i-f 94. Abu further teaches that nodes "may be placed within
the [grid] hierarchy or removed from the hierarchy while the distributed task
is in progress, and the new node allocated a portion of the distributed task."
Id. i-f 125.
In view of these disclosures, we are not persuaded by Appellants'
argument that Abu's teaching is limited to allocating tasks in a
predetermined and fixed manner, and that Abu fails to teach or suggest
dynamically splitting a job or task that had been allocated to a first node into
multiple jobs or tasks that are allocated to different nodes as required by
claim 1. "The test for obviousness is not ... that the claimed invention must
be expressly suggested in any one or all of the references. Rather, the test is
what the combined teachings of the references would have suggested to
those of ordinary skill in the art." In re Keller, 642 F.2d 413, 425 (CCPA
1981 ).
4
Appeal2015-000104
Application 12/821,784
Issue 2: Whether Abu teaches or suggests that an interprocess
communication between processing units occurs within a node
before splitting a job and over a communication link between
different nodes after splitting the job as required by claim 1.
The Examiner finds Abu teaches or suggests this limitation. Final
Act. 5 (citing Abu i-fi-150, 61, 62, 67, 71, 72, and 122-125); Ans. 11-12.
Appellants argue the portions of Abu cited by the Examiner "describe the
super-node and how the communication between nodes is handled by the
self organizing grid system," and that "[n]othing in these cited sections
describe [that] interprocess communication between processing units of the
job is within a node before splitting the job," and is handled "over a
communication link between the different nodes" after splitting the job.
App. Br. 11-12. We are not persuaded by Appellants' arguments.
As noted supra, "[t]he test for obviousness is not ... that the claimed
invention must be expressly suggested in any one or all of the references.
Rather, the test is what the combined teachings of the references would have
suggested to those of ordinary skill in the art." In re Keller, 642 F.2d at 425.
Appellants' "interprocess communication" limitation is little more than a
tautological statement that the communications between two processing
units occurs between the locations where the two processing units are
executing. That is, when the two processing units are executing within a
node, they communicate between locations within the node. When the job is
split, and the two processing units are executing on different nodes, they
communicate over a link between the locations of the different nodes. Abu
teaches or suggests this tautological truth. For example, Abu teaches that
when two tasks are executing on nodes in the same neighborhood (e.g.,
5
Appeal2015-000104
Application 12/821,784
nodes 2 and 6 in layer 0 of Figure 1 ), they communicate over links within
the neighborhood. However, when the tasks are executing on nodes in
different neighborhoods (e.g., nodes 4 and 6 in layer 0 of Figure 1 ), they
communicate through their respective supemodes (e.g., nodes 2 and 15 in
layer 1 of Figure 1) over links between the neighborhoods. See Abu i-f 62,
Fig. 1. Thus, we find the sections of Abu cited by the Examiner teach or
suggest the "interprocess communication" limitations to a person of ordinary
skill in the art.
Accordingly, for the reasons discussed above, we sustain the
Examiner's rejection of claim 1, as well as of claims 2, 4, 8, 16, 17, 21, and
22, not separately argued. See App. Br. 7-13, 15.
Claims 5-7, 19, and 20
Issue 3: Whether the combination of Abu and Votta teaches or
11 • • • ,] ., 1 • • suggests COllectzng rnetrzcs assoczateu wztrz trze processzng unzts,
application, and nodes, where the metrics are selected from CPU
utilization, memory utilization, throughput, latency, and heap
size.
Claims 5-7, 19, and 20 require the collected metrics to be associated
with the processing units (claims 5/19), the application (claims 6/20), and
the nodes on which the processing units are executing (claim 7). The claims
also require the collected metrics to be selected from CPU utilization (all
claims), memory utilization (all claims), throughput (claims 5/19 and 6/20),
latency (claims 5/19 and 6/20), and heap size (claim 7).
The Examiner finds the combination of Abu and Votta teaches or
suggests these limitations. Final Act. 11-13, 15. In particular, the Examiner
6
Appeal2015-000104
Application 12/821,784
finds Abu teaches collecting resource utilization metrics, and Votta teaches
resource utilization metrics can include CPU utilization, memory utilization,
throughput, and latency. Id. (citing Abu i-fi-130, 39, 93; Votta i153; Fig. 3).
Appellants argue the Examiner erred in rejecting claims 5-7, 19, and
20 because Votta' s metrics are general CPU utilization metrics rather than
metrics for "CPU utilization by the processing unit" as recited in claims
5/19, metrics for "CPU utilization by the application" as recited in claims
6120, and metrics for "CPU utilization for the node" as recited in claim 7.
App. Br. 14, 16. We are not persuaded by Appellants' arguments.
Abu teaches allocating system resources to optimize the performance
of the grid as a whole, and individual applications running on the grid. Abu
i130. Abu further teaches that effective resource allocation requires
considering different metrics and constraints, such as resource utilization for
different jobs, and for different storage, computing, and network resources.
Id. i193. Abu further teaches that peak performance can be achieved by
minimizing job execution time and communication delays, and maximizing
resource utilization. Id. i194. Votta teaches enhancing throughput in grid
computing systems based on monitored software parameters including, but
not limited to, the load on one or more processors, CPU utilization, memory
utilization, transaction latency, system throughput, and other monitored
performance metrics gathered from the operating system. Votta i-fi-1 3, 5 3.
As noted supra, "[t]he test for obviousness is not ... that the claimed
invention must be expressly suggested in any one or all of the references.
Rather, the test is what the combined teachings of the references would have
suggested to those of ordinary skill in the art." In re Keller, 642 F.2d at 425.
Appellants' claims 5-7, 19, and 20 recite collecting various performance
7
Appeal2015-000104
Application 12/821,784
metrics at different levels of granularity, namely, from individual tasks or
processing units, from the application as a whole (i.e., from the sum of the
processing units), and from the sum of the nodes on which the processing
units, and therefore application, are executing. Votta similarly teaches or
suggests gathering performance metrics at different levels of granularity
based on parameters obtained from the operating system (e.g., the load on
one or more processors, individual transaction latencies, and system
throughput). Accordingly, we are not persuaded by Appellants' argument
that the combination of Abu and Votta fails to teach or suggest gathering
various performance metrics (e.g., CPU utilization, latency, throughput) for
the processing units (claims 5/19), for the application (claims 6/20), or for
the nodes (claim 7).
Accordingly, for the reasons discussed, we sustain the Examiner's
rejections of claims 5-7, 19, and 20.
DECISION
The rejection of claims 1, 16, and 21under35 U.S.C. § 103(a) as
unpatentable over Abu-Ghazaleh and Fellenstein is affirmed.
The rejection of claims 2 and 17 under 35 U.S.C. § 103(a) as
unpatentable over Abu-Ghazaleh, Fellenstein, and Barsness is affirmed.
The rejection of claim 4 under 35 U.S.C. § 103(a) as unpatentable
over Abu-Ghazaleh, Fellenstein, and Dillenberger is affirmed.
The rejection of claims 5, 6, 8, 19, 20, and 22 under 35 U.S.C.
§ 103(a) as unpatentable over Abu-Ghazaleh, Fellenstein, and Votta is
affirmed.
8
Appeal2015-000104
Application 12/821,784
The rejection of claim 7 under 35 U.S.C. § 103(a) as unpatentable
over Abu-Ghazaleh, Fellenstein, Votta, and Kaminsky is affirmed.
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
9