Ex Parte Saito et al

Board of Patent Appeals and Interferences

Feb 6, 2012

10726041 (B.P.A.I. Feb. 6, 2012)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
____________

Ex parte YASUSHI SAITO and CHRISTOS KARAMANOLIS
____________

Appeal 2009-005395
Application 10/726,041
Technology Center 2100
____________

Before JOHN A. JEFFERY, MARC S. HOFF, and BRADLEY W.
BAUMEISTER, Administrative Patent Judges.

JEFFERY, Administrative Patent Judge.

DECISION ON APPEAL
Appellants appeal under 35 U.S.C. § 134(a) from the Examiner’s
rejection of claims 1-15 and 56-69. We have jurisdiction under 35 U.S.C.
§ 6(b). We reverse.

STATEMENT OF THE CASE
Appellants’ invention maintains namespace consistency in a wide-
area file system by propagating file replica updates to replicas at other
nodes. See generally Abstract; Spec. 27-28; Fig. 2. Claim 1 is illustrative:
Appeal 2009-005395
Application 10/726,041

2
1. A method for a wide-area file system, including a plurality of nodes
storing replicas of objects, the objects being files and file directories,
wherein for each replica of an object at a node, a parent directory for the
object is replicated at the node, the method comprising:

propagating an update to a replica of a file directory to other replicas
of the file directory via a graph, wherein each replica of the file directory has
edges to only a subset of the other replicas such that all the replicas of the
file directory are connected via the graph; and

in response to receiving a propagated update to a replica of the file
directory at a node, updating the replica for the file directory at the node.

The Examiner relies on the following as evidence of unpatentability:
Moulton US 2001/0044879 A1 Nov. 22, 2001
Coram US 2002/0107835 A1 Aug. 8, 2002
Gerald J. Popek et al., Replication in Ficus Distributed File Systems, Proc.
of the Workshop on Mgt. of Replicated Data, Nov. 1990, at 20-25 (“Popek”).

Zheng Zhang & Christos Karamanolis, Designing a Robust Namespace for
Distributed File Services, Hewlett-Packard Laboratories, Oct. 2001
(“Zhang”).

Haifeng Yu & Amin Vahdat, The Costs and Limits of Availability for
Replicated Services, Proc. of the 18th ACM Symp. On Operating System
Principles, Oct. 2001 (“Yu”).

Appeal 2009-005395
Application 10/726,041

3
THE REJECTIONS1
1. The Examiner rejected claims 1, 10, and 11 under 35 U.S.C.
§ 103(a) as unpatentable over Popek and Moulton. Ans. 4-6.2
2. The Examiner rejected claims 2-9 and 56-66 under 35 U.S.C.
§ 103(a) as unpatentable over Popek, Moulton, and Zhang. Ans. 6-11.
3. The Examiner rejected claim 12 under 35 U.S.C. § 103(a) as
unpatentable over Popek, Moulton, and Yu. Ans. 12-13.
4. The Examiner rejected claims 13 and 15 under 35 U.S.C. § 103(a)
as unpatentable over Popek, Moulton, and Coram. Ans. 13-14.
5. The Examiner rejected claims 14 and 67-69 under 35 U.S.C.
§ 103(a) as unpatentable over Popek, Moulton, Coram, and Zhang. Ans. 14-
16.

THE OBVIOUSNESS REJECTION OVER POPEK AND MOULTON
The Examiner finds that Popek’s wide-area file system has every
recited feature of independent claim 1 except for propagating file directory
replica updates via a graph, where each replica has edges to only a subset of
the other replicas such that all file directory replicas are connected via the

1 Since the Examiner withdrew rejections under § 112 (Ans. 3), they are not
before us. We likewise do not consider Appellants’ arguments regarding the
drawing objection (App. Br. 12) since they pertain to a petitionable—not
appealable—matter as the Examiner indicates. Ans. 2.
2 Throughout this opinion, we refer to (1) the Appeal Brief filed March 7,
2008; (2) the Examiner’s Answer mailed May 8, 2008; and (3) the Reply
Brief filed July 7, 2008.
Appeal 2009-005395
Application 10/726,041

4
graph. The Examiner, however, cites Moulton as teaching this feature in
concluding that the claim would have been obvious. Ans. 4-6, 16-20.
Appellants argue that is it not obvious to combine the references as
proposed since, among other things, Popek’s distributed file system is
inapplicable to Moulton’s system, which stores divided data elements across
multiple nodes. App. Br. 16-19; Reply Br. 4-7. Appellants add that even if
combined, the cited references do not teach or suggest replicating a parent
directory for each object replica at a node, let alone propagating file
directory replica updates via a graph, where each replica has edges to only a
subset of the other replicas such that all file directory replicas are connected
via the graph. App. Br. 19-21; Reply Br. 8-10. The issue before us, then, is
as follows:

ISSUE
Under § 103, has the Examiner erred in rejecting claim 1 by finding
that Popek and Moulton collectively would have taught or suggested plural
nodes storing replicas of objects, the objects being files and file directories,
wherein for each object replica at a node, a parent directory for the object is
replicated at the node?

FINDINGS OF FACT (FF)
1. “An ‘edge’ represents a known connection between two replicas of
a file; updates to the file flow along edges. The replicas of a file and the
Appeal 2009-005395
Application 10/726,041

5
edges between the replicas comprise a strongly connected ‘graph’.” Spec.
22:15-19.
2. Popek notes that a key issue with replication in a distributed
environment is mutual consistency (i.e., keeping multiple copies of an object
consistent with each other). To this end, one approach involves updating in
one connected environment, and then propagating that update to other
storage sites. Popek, at 21, § 3 (“The Optimistic Model”).
3. Popek’s system can replicate files selectively. A collection of file
volume replicas are set up at various storage sites for a given logical file
subtree. A given file may be replicated at any subset of the sites hosting a
volume replica. Popek, at 22, § 5 (“The Fiscus Project”).
4. Moulton stores data in an internetwork environment 101 (e.g., the
Internet) connecting multiple Wide Area Networks 103 and Local Area
Networks (LANs) 104. The LANs are associated with storage mechanisms
106 that control storage devices at nodes 105. Moulton, ¶¶ 0032, 0036; Fig.
1. A partial detail view of Moulton’s Figure 1 showing LANs and
associated storage mechanisms and nodes is shown below:

Appeal 2009-005395
Application 10/726,041

6

Partial detail view of Moulton’s Figure 1 showing LANs and associated
storage mechanisms and nodes

5. Moulton’s Figures 7A-F show various redundant array of
inexpensive nodes (RAIN) protection levels that include data striping (Figs.
7A, 7C-E) and mirroring (Fig. 7B). Moulton, ¶¶ 0067-86; Figs. 7A-F.
6. A backpointer need not remember the locations of parent-directory
replicas because a parent directory is always found on the same node. Spec.
27:18-22.

ANALYSIS
Based on the record before us, we find error in the Examiner’s
obviousness rejection of independent claim 1, which recites, in pertinent
part, plural nodes storing replicas of objects, the objects being files and file
directories, wherein for each object replica at a node, a parent directory for
the object is replicated at the node.
Appeal 2009-005395
Application 10/726,041

7
To be sure, Popek replicates file volumes at various storage sites,
which at least suggests replicating directories at plural nodes consistent with
the Examiner’s mapping. FF 3; Ans. 18-19. And by selectively replicating
files associated with volume replicas at particular storage sites, Popek at
least suggests propagating updates to those sites. See FF 2-3. But we cannot
say—nor has the Examiner shown—that the cited prior art teaches or
suggests for each replica of a file object and directory object at a node,
replicating a parent directory at the node as claimed. As Appellants
indicate, this “namespace containment” process improves availability and
administration when propagating replica updates among plural nodes by,
among other things, replicating all intermediate pathname components on
the same node. See App. Br. 19-20 (referring to Spec. 15-16); Reply Br. 8.
Namespace containment is particularly crucial to Appellants’ invention
since, among other things, a backpointer (see, e.g., backpointer 225 in Fig.
2) need not remember the locations of parent-directory replicas because a
parent directory is always found on the same node (FF 6)—a feature that
simplifies propagating object replica updates.
Popek’s file volume replicas are established at various storage sites
for a given logical file subtree (FF 3), which, at best, merely suggests
replicating directories in a hierarchy. But even assuming, without deciding,
that some of those directories are parent directories, we cannot say—nor has
the Examiner shown—that the cited prior art teaches replicating a parent
directory for each file and directory replica at each node as claimed.
Appeal 2009-005395
Application 10/726,041

8
As the Examiner concedes (Ans. 19), Popek does not specify exactly
how updates are propagated, let alone using the recited technique. To cure
this acknowledged deficiency, the Examiner cites Moulton’s internet-based
network configuration for allegedly teaching that each file directory replica
in a particular node (e.g., node 105 in Moulton) has edges to only a subset of
other replicas (e.g., other nodes 105 or the storage mechanism 106) via LAN
“cloud” 104. Ans. 20. The Examiner adds that all file directory replicas are
“eventually connected” through the clouds in this “graph.” Id.
Even assuming that (1) Moulton’s LANs include “edges” under
Appellants’ definition of the term by connecting two nodes within the LAN
(FF 1, 4), and (2) these nodes contain replicated directories as in Popek (FF
3) to collectively constitute a “graph,” we still cannot say that this topology
reasonably teaches or suggests replicating a parent directory for each file
and directory replica at each node as claimed. Although we find the
references reasonably combinable in view of Moulton’s mirroring (i.e.,
replicating) data between storage devices as the Examiner indicates (Ans.
17; FF 5), the cited prior art nonetheless collectively fails to teach or suggest
the limitations of claim 1.
We are therefore persuaded that the Examiner erred in rejecting
independent claim 1 and dependent claims 10 and 11 for similar reasons.
Since this issue is dispositive regarding our reversing the rejection of these
claims, we need not address Appellants’ other arguments pertaining to these
claims (App. Br. 21-22; Reply Br. 10-13).
Appeal 2009-005395
Application 10/726,041

9
THE OBVIOUSNESS REJECTION OVER POPEK, MOULTON, AND ZHANG
Since the Examiner has not shown that Zhang cures the deficiencies
noted above regarding independent claim 1 and commensurate limitations in
independent claim 56, we reverse the obviousness rejection of claims 2-9
and 56-66 (Ans. 6-11) for the reasons noted above. Since this issue is
dispositive regarding our reversing the rejection of these claims, we need not
address Appellants’ other arguments pertaining to this rejection (App. Br.
22-34; Reply Br. 13-15).

THE OTHER OBVIOUSNESS REJECTIONS
Since the Examiner has not shown that the other cited prior art cures
the deficiencies noted above regarding independent claims 1 and 56, we
reverse the obviousness rejections of dependent claims 12-15 and 67-69
(Ans. 12-16) for the reasons noted above.

CONCLUSION
The Examiner erred in rejecting claims 1-15 and 56-69 under § 103.

ORDER
The Examiner’s decision rejecting claims 1-15 and 56-69 is reversed.

REVERSED
rwk