Ex Parte Desai et al

Patent Trial and Appeal Board

May 13, 2014

11180072 (P.T.A.B. May. 13, 2014)

UNITED STATES PATENT AND TRADEMARK OFFICE 1
___________ 2
3
BEFORE THE PATENT TRIAL AND APPEAL BOARD 4
___________ 5
6
Ex parte ADITYA A. DESAI, BRIAN K. MARTIN, 7
JASON MCGEE, and GABRIEL G. MONTERO 8
___________ 9
10
Appeal 2011-009372 11
Application 11/180,072 12
Technology Center 2400 13
___________ 14
15
16
Before HUBERT C. LORIN, ANTON W. FETTING, and 17
THOMAS F. SMEGAL, Administrative Patent Judges. 18
19
FETTING, Administrative Patent Judge. 20
21
22
DECISION ON APPEAL 23
Appeal 2011-009372
Application 11/180,072

2
STATEMENT OF THE CASE1 1

1 Our decision will make reference to the Appellants’ Appeal Brief (“App.
Br.,” filed November 26, 2010) and the Examiner’s Answer (“Ans.,” mailed
February 18, 2011).
Aditya A. Desai, Brian K. Martin, Jason McGee, and Gabriel G. 2
Montero (Appellants) seek review under 35 U.S.C. § 134 of a final rejection 3
of claims 1-17, the only claims pending in the application on appeal. We 4
have jurisdiction over the appeal pursuant to 35 U.S.C. § 6(b). 5
The Appellants invented a way of maintaining server affinity while 6
load balancing requests to a server cluster that rebalances sessions across a 7
server cluster having multiple routers without coordinating the multiple 8
routers and while accounting for configuration changes in the server cluster 9
(Specification para. [0001]). 10
An understanding of the invention can be derived from a reading of 11
exemplary claim 6, which is reproduced below [bracketed matter and some 12
paragraphing added]. 13
6. A computer implemented method comprising: 14
[1] using a filter program 15
residing in a storage connected to a server cluster, 16
intercepting a client request 17
in a session from a client computer 18
directed to a first server in the server cluster 19
by one router of a plurality of routers 20
and 21
sending a filter request to an analysis program, 22
each router in the plurality of routers distributes client 23
requests to the server cluster; 24
[2] responsive to receiving instructions from the analysis 25
program, 26
adding a cookie to the client request 27
after establishing the session, 28
Appeal 2011-009372
Application 11/180,072

3
and 1
sending the client request, 2
including the cookie, 3
to a second server 4
chosen by the analysis component 5
in order to achieve a rebalance of a workload 6
within the server cluster 7
during the session; 8
[3] wherein the cookie contains a redirect command 9
instructing the client computer 10
to send all subsequent client request in the session 11
to the second server; 12
[4] wherein the filter program continually redirects 13
each subsequent client request within the session 14
to the second server 15
until 16
the client computer receives the redirect command 17
and 18
begins to send any post command client requests in 19
the session to the second server; 20
and 21
[5] wherein the session is moved 22
from the first server 23
to the second server 24
to achieve the rebalance of the workload 25
independently of the plurality of routers. 26
The Examiner relies upon the following prior art: 27
Choquier US 5,951,694 Sep. 14, 1999
Amano US 2003/0187871 A1 Oct. 2, 2003
28
Claims 1-17 stand rejected under 35 U.S.C. § 103(a) as unpatentable 29
over Choquier and Amano. 30
Appeal 2011-009372
Application 11/180,072

4
ISSUES 1
The issues of obviousness turn primarily on whether there is a reason 2
to place Choquier’s load balancing program on all of its servers and whether 3
the claims recite session ID’s in the limitations. 4
5
FACTS PERTINENT TO THE ISSUES 6
The following enumerated Findings of Fact (FF) are believed to be 7
supported by a preponderance of the evidence. 8
Facts Related to the Prior Art 9
Choquier 10
01. Choquier is directed to large-scale computer networks for 11
providing information-related services and communications 12
services to end users. Choquier 1:15-17. 13
02. Multiple Gateway microcomputers are connected to the local 14
area network. Each Gateway microcomputer is for 15
communicating with client microcomputers of end users. The 16
Gateway microcomputers serve as intermediaries between the 17
client microcomputers and the application servers. When a user 18
initially logs onto the network, a connection is established 19
between the client microcomputer and a Gateway microcomputer, 20
and the connection is maintained throughout the logon session. 21
During the logon session, the user typically opens and closes 22
multiple services, and thus communicates with multiple 23
application servers. The Gateway microcomputer assigned to the 24
logon session maintains the various client-server message streams 25
by routing service requests (from the client microcomputer) to the 26
Appeal 2011-009372
Application 11/180,072

5
appropriate application servers, and by returning application 1
server responses to the client microcomputer. Choquier 2:26-47. 2
03. In accordance with a dynamic load balancing feature, when a 3
user sends a request to open a service, the Gateway 4
microcomputer that receives the request initially identifies the 5
application servers that are within the relevant service group. The 6
Gateway microcomputer then determines the current load of each 7
application server in the service group, and applies a load 8
balancing method to select an application server that is relatively 9
lightly loaded. The service request is then passed to the selected 10
application server for processing. Choquier 2:48-57. 11
04. The Gateway microcomputers identify the application servers 12
in the service groups and determine the loads of the application 13
servers by accessing a locally-stored service map. The service 14
map contains information about every application server of the 15
system. For example, for each application server, the service map 16
indicates the names of the server applications running on the 17
application server, the current processing load of the application 18
server, and the overall processing power (in the form of a static 19
CPU INDEX benchmark) of the application server. The service 20
map is updated on the order of every 30 seconds to reflect changes 21
in application server loads, and to reflect various changes in the 22
system configuration. Choquier 2:58 – 3:3. 23
05. Load balancing is preferably performed according to the 24
specific dynamics of the individual services. One load balancing 25
method uses the load and CPU INDEX values contained in the 26
Appeal 2011-009372
Application 11/180,072

6
service map to calculate the available CPU processing power for 1
each application server in the relevant service-group, and then 2
passes the service request to the application server having the 3
greatest available CPU processing power. Another load balancing 4
method selects application servers such that the probability of 5
selection of a given application server is proportional to that 6
application server’s available CPU processing power. Choquier 7
3:4-14. 8
06. Application servers may be dynamically allocated to services to 9
accommodate for fluctuations in usage levels of different services. 10
To allocate an application server to a particular service, the server 11
application that implements the service is dynamically loaded (i.e., 12
loaded without stopping the underlying process) on the application 13
server. Likewise, to remove an application server from a service 14
group, the relevant server applications are dynamically unloaded. 15
Allocation of application servers to service groups can be 16
performed semi-automatically by a system operator who monitors 17
service loads from a system console, or automatically using a load 18
monitoring program that compares service loads to predetermined 19
thresholds. Choquier 3:32-50. 20
07. Application servers may be dynamically allocated to services to 21
accommodate for fluctuations in usage levels of different services. 22
To allocate an application server to a particular service, the server 23
application that implements the service is dynamically loaded (i.e., 24
loaded without stopping the underlying process) on the application 25
server. Likewise, to remove an application server from a service 26
Appeal 2011-009372
Application 11/180,072

7
group, the relevant server applications are dynamically unloaded. 1
Allocation of application servers to service groups can be 2
performed semi-automatically by a system operator who monitors 3
service loads from a system console, or automatically using a load 4
monitoring program that compares service loads to predetermined 5
thresholds. Choquier 3:32-50. 6
08. FIG. 6 illustrates a preferred sequence of steps that is performed 7
whenever a Gateway receives an "open" request from a client-8
user. Gateway initially accesses the current service map to 9
determine the number of servers that are allocated to the requested 10
service. If only one server is currently assigned to the service, the 11
Gateway routes the service request to that server. Choquier 13:56-12
65. 13
09. If multiple servers are assigned to the service group, the 14
Gateway then determines whether or not the particular service 15
requires load balancing. Although load balancing is appropriate 16
for most services, certain services require that service requests be 17
handled by a particular server, and thus require the use of a 18
custom locator program to identify the correct server. Choquier 19
13:66 – 14:6. 20
10. If the requested service does not require the use of a custom 21
locator program, the Gateway reads the service map to obtain the 22
CPU LOAD and CPU INDEX values for every server within the 23
service group. The Gateway then accesses a table which specifies 24
a load balancing method to use for the service. Although the same 25
load balancing method could be used for all services, it has been 26
Appeal 2011-009372
Application 11/180,072

8
found that a more uniform load balancing result can be achieved 1
by taking into account the differing processing demands placed on 2
the servers by different service applications. The "open" request 3
is then passed to the selected server. Choquier 13:66 – 14:6. 4
11. Various alternative load balancing methods are possible, and 5
the load balancing function need not be performed by the 6
Gateways. For example, the Gateways could be programmed to 7
pass each “open” request to one of the servers within the service 8
group, and the recipient server could apply the load balancing 9
method and (if necessary) reroute the request. Choquier 16:1-8. 10
12. Prior to dynamically unloading a service DLL from a server, all 11
sessions of the service (if any) must be moved to another server 12
that is handling the same service. A “hot redirection” technique is 13
used to transfer service sessions. The hot redirection technique 14
works as follows. Initially, the Gateway that is handling the logon 15
session submits a serialization request to the server that is 16
handling the service session. The server responds to the 17
serialization request by returning an object that describes the 18
internal state of the service session. This object is in the form of a 19
byte stream which is normally meaningful only to a specific 20
service. By way of example, the object for the hot redirection of a 21
CHAT service session would normally indicate the conference 22
name, the user name, the status of the user (e.g., a spectator, 23
moderator, etc.), and the message number of the last CHAT 24
message sent to the user in the CHAT conference. While the 25
server is responding to the serialization request, the Gateway 26
Appeal 2011-009372
Application 11/180,072

9
buffers any service requests that are received from the client 1
microcomputer. The Gateway then forwards the internal state 2
information and the buffered service requests (if any) to a new 3
server (using the load balancing procedure of FIG. 6 to select a 4
new server), and the new server resumes processing of service 5
requests where the former server left off. Once all service 6
sessions have been transferred in this manner, the service DLL can 7
be dynamically unloaded. A new service DLL can then be 8
dynamically, loaded to activate a different service, or to activate 9
an updated version of the removed service. The above-described 10
hot redirection technique improves service reliability from the 11
perspective of the end user by allowing service sessions to 12
continue when corresponding servers are taken down for 13
maintenance. It is contemplated that this technique will be 14
combined with a fault detection mechanism to automatically 15
transfer service sessions to different servers when signs of 16
abnormal service behavior are present. Choquier 17:57 – 18:23. 17
13. Hot redirection enables a service session to be transferred from 18
one server to another, so that a server can be removed from a 19
service group without prematurely terminating a service session 20
Choquier 23:60-63. 21
Amano 22
14. Amano is directed to handling session information for use on a 23
Web by inheriting the session information when a browser side is 24
cooperatively directed from an original server site to another site 25
and then restored to the original site. Amano para. [0001]. 26
Appeal 2011-009372
Application 11/180,072

10
15. Amano describes a first redirect application for accepting a first 1
request for session and a second redirect application for accepting 2
a request in getting back from another site. The first redirect 3
application and the second redirect application make a redirect 4
process of returning a response to the browser side once, and 5
automatically transmitting a request to the server side again, and 6
return a response to the browser side. The browser side 7
automatically transmits a request to the real application on the 8
server side again, in accordance with a description in the response, 9
without making the user aware of it. Amano para. [0019]. 10
16. Amano describes inheriting the session information, including a 11
step of receiving an HTTP request, a step of reading the cluster 12
information to be redirected from a configuration file of an 13
operating server, a step of generating a redirect response for 14
initiating the real application employing the parameters received 15
upon the HTTP request and the cluster information, a step of 16
setting the cluster information as a cookie in the redirect 17
response, a step of transmitting the redirect response to the 18
browser side, a step of receiving a new HTTP request from the 19
browser side piloted from the other site, a step of acquiring the 20
cluster information that is set in the new HTTP request, a step of 21
generating a new redirect response employing the parameters 22
received upon the new HTTP request and the acquired cluster 23
information, and a step of transmitting the new redirect response 24
to the browser side. Amano also describes inheriting the session 25
information, including a step of checking whether or not the 26
Appeal 2011-009372
Application 11/180,072

11
cluster information is embedded in a cookie for a received HTTP 1
request, a step of setting the cluster information or the read 2
cluster information as the cookie in the redirect response, and a 3
step of transmitting the redirect response. Amano para. [0021]. 4
17. It is known to use cookies to maintain a “stick” session to a 5
server in a cluster after an initial step of selecting the server by 6
load balancing. Amano para. [0044]. 7
18. The redirect application A sets the cluster information read 8
from the configuration file as a cookie in a Set-Cookie header 9
within the header of the redirect response. Amano para. [0054]. 10
ANALYSIS 11
As to claims 1 and 12, reciting “a filter program residing in a storage 12
in each of the servers of the server cluster,” we are persuaded by the 13
Appellants’ argument that these claims require all servers in the plurality 14
(more than one) of servers to contain the filter program. Br. 8. The 15
Examiner finds that Choquier describes allowing a server other than its 16
Gateway server to apply the equivalent of the recited filter program. The 17
Examiner in particular finds 18
Choquier discloses at column 16 lines 1-8 the load 19
balancing function need not be performed by the Gateways. 20
The Gateways could be programmed to pass each request to one 21
of the servers within the service group, and the recipient server 22
could apply the load balancing method and reroute the request. 23
As such Choquier discloses that the functionality of the 24
Gateways may instead be performed by the servers. 25
Ans. 24. 26
This finding conflates capacity with motivation. Having the filter 27
program in one of the servers does not in itself suggest having that same 28
Appeal 2011-009372
Application 11/180,072

12
program in all servers. Choquier uses a map of servers for the balancing 1
program to consult. With such a map that a single server may consult, it is 2
not seen why Choquier would replicate the balancing program across all 3
servers, and Examiner has not articulated any reason for doing so. The 4
Examiner finds that there was a reason to copy Choquier’s hot redirection 5
across all servers (Ans. 5), but as Appellants contend at Brief 9, there is no 6
reason to apply a cookie with a hot redirection, as the process is transparent 7
to the client. Also, Examiner’s finding that Choquier’s hot redirection relies 8
on Choquier’s load balancing technique does not imply that the balancing 9
algorithm is replicated to each server, but only that Choquier’s load 10
balancing server is consulted. 11
As to claims 6 and 9, reciting “responsive to receiving instructions 12
from the analysis program, adding a cookie to the client request after 13
establishing the session, and sending the client request, including the cookie, 14
to a second server chosen by the analysis component in order to achieve a 15
rebalance of a workload within the server cluster during the session,” we are 16
not persuaded by the Appellants’ argument that Amano does not show 17
redirecting client requests during a session from one application server to 18
another, where the cookie is included and transmitted to the second 19
application server. Appellants contend that Amano’s redirect response 20
occurs before a session object is created, and before a session ID is set in the 21
cookie. Br. 11. Appellants further contend that Amano is 22
using the cluster address rather than the previously assigned 23
server address for a particular server, the information in the 24
cookie allows the client's requests upon returning to the server 25
cluster to continue to be handled by the originally selected 26
application server so that the session is maintained. In other 27
words, after the session is established, even when the browser 28
Appeal 2011-009372
Application 11/180,072

13
“pilots away” from the web site (“server cluster”) it is not 1
redirected to another server, rather the server affiliation 2
established in the session initialization is maintained. 3
Id. 4
5
As to the contention regarding a session object and ID set in the 6
cookie, this is not commensurate with the scope of the claim. As to the 7
contention regarding session timing, the claim provides no limitation on the 8
characteristics or definition of the recited session, and so Amano’s timing is 9
within the scope of a generic recitation of a session. 10
As to the contention regarding redirection, limitation [4] does not 11
further limit the manner of redirection. The manner of redirection is set 12
forth in limitation [3], being instructing the client computer to send all send 13
all subsequent client request in the session to the second server. This is 14
precisely what Amano does. 15
CONCLUSIONS OF LAW 16
The rejection of claims 6-11 under 35 U.S.C. § 103(a) as unpatentable 17
over Choquier and Amano is proper. 18
The rejection of claims 1-5 and 12-17 under 35 U.S.C. § 103(a) as 19
unpatentable over Choquier and Amano is improper. 20
DECISION 21
The rejection of claims 6-11 is affirmed. 22
The rejection of claims 1-5 and 12-17 is reversed. 23
24
Appeal 2011-009372
Application 11/180,072

14
No time period for taking any subsequent action in connection with 1
this appeal may be extended under 37 C.F.R. § 1.136(a). See 37 C.F.R. 2
§ 1.136(a)(1)(iv) (2011). 3
4
AFFIRMED-IN-PART 5
6
Klh 7