HP Integrity NonStop H-Series - Manual (Page 250)

4. Set the value of TANDEM_RECEIVE_DEPTH to 250 (because the values of Maxlinks and
TANDEM_RECEIVE_DEPTH should be equal). A value of 250 means that the NSJSP process
is able to read a maximum of 250 messages simultaneously from its $RECEIVE file.
5. Configure the Executor element in <NSJSP deployment
directory>/conf/server.xml file on OSS. The Executor element is used by the
Connector element as a thread pool. Set maxThreads = 300 , minSpareThreads =
10 and maxIdleTime = 30000 . Using an Executor element helps to monitor the number
of active threads at any given time. A value of 300 for maxThreads ensures that you have
enough threads to process all the incoming requests (a maximum of 250) simultaneously.
A value of 30000 for
maxIdleTime ensures that if a thread is idle for more than 30 seconds,
that thread will be stopped.
After making these configurations, you might want to use a tool that can simulate the HTTP
clients and can handle the HTTP cookies. The tool reveals the number of HTTP clients that
one single instance of NSJSP can handle and indicates the number of simultaneous HTTP
requests that NSJSP is capable of handling.
The following are the suggested steps to arrive at the required numbers:
a. Run the test tool with a single instance of the HTTP client simulator.
b. The test tool must be capable of displaying the response time of the HTTP requests.
Monitor the response time of the HTTP requests and allow your application to attain
a steady state (in terms of response time per HTTP Request).
c. At steady state, check if the response times are within the Service Level Agreement
(SLA).
d. If the response time is within the SLA, increase the number of HTTP client instances
and repeat step 2 onwards.
e. If the response time is beyond the acceptable SLA, stop the tests and determine the
number of HTTP clients and the number of HTTP requests that the NSJSP instance can
process simultaneously.
While the test tool can indicate the number of HTTP clients that are processed
simultaneously, the number of simultaneous HTTP requests can be arrived at using
different means. Following are some of the calculation strategies:
• The number of HTTP requests recorded by the testing tool and idea of the number
of requests, which required processing by NSJSP, provides the total number of
HTTP requests processed by NSJSP.
• The total number of HTTP requests processed by NSJSP together with the test
duration indicates the average number of simultaneous HTTP requests handled
by NSJSP.
NOTE: There are a number of HTTP client simulators available, for example,
Apache JMeter , Mercury LoadRunner , and Radview Webload . These tools
provide a good interface to monitor the test results. You can use any of these tools
for determining the maximum handling capacity of an NSJSP instance.
250 Configuring MyFaces Applications on NonStop Systems