Hi All,

currently we are using 11.1.R6 @Linux Environment and we are having performance issues and we are planning to upgrade to 11.1.R7. what is the best practices/ways to improve the performance of the server. currently we are using one gateway and 2 dispatchers. and can we configure separate gateways for Reporting and Schedule/Job services for Cognos 11.1.R7. Can any one suggest me what is the best practices to improve the performance and Gateway Settings.

Hello Srikanth,

Not sure what your performance issues are. Running reports, dashboards, navigating the portal, administration, schedules... 
There are many areas where the perception of performance has nothing to do with the architecture. 

I would encourage you to investigate what the issues are rather than trying to answer the question with more hardware or tuning. If your car has square wheels, adding a race car engine will not help. 

Cognos Analytics does not need a gateway. You might choose to use one for SSO and image hosting but you certainly don't need separate gateways for reporting or scheduling. If you are considering resources for batch reporting and interactive reporting I would encourage you to read the tuning documentation. Business reporting requirements and patterns change over time and ongoing tuning may be required in order to allocate appropriate resources. Tuning server performance (ibm.com)

You might like to consider using IBM services to review application health, provide best practice guidance, and tune for performance. 

------------------------------
Luke Peterson
Business Analytics Advocate
IBM
------------------------------

Without a deeper understanding of your processing issue - You could set up routing rules to force particular jobs ETC. to a given server.  forcing your heavier jobs to one server leaving the other for day to day jobs - a second gateway is possible but that would require making the user use that particular address. routing rules would be my first choice

​

Hello Srikanth,

Depending on where the performance is being lost you might want to add power to DB, APP or GW.

Luke mentioned something similiar. Routing blocking requests (schedules) is always a good idea.

It would be nice to understand a little bit more on where the time "is lost" or the performance bottleneck is suspected.

Load Balancing on GW + APP

if the GW is the chokepoint (which I doubt!!), then adding a LoadBalancer (Hardware or Software, e.g. haproxy or Apache LB Module) in Front of the GW and adding more GW is the thing to go for.

See GW balancing: https://httpd.apache.org/docs/2.4/mod/mod_proxy_balancer.html

See APP Server Balancing - https://www.ibm.com/support/knowledgecenter/en/SSEP7J_11.0.0/com.ibm.swg.ba.cognos.inst_cr_winux.doc/inst_cr_winux.pdf, Page 91 ff

#Set up a cluster for load-balancing
# Include all Cognos Analytics servers that have the Application server components
# installed and configured.
# Important: do not include Cognos Analytics servers that are used to run the
# Content Manager service. 

Check browser if static content is delivered from client:

Retrieving files from GW and App Server speeds up when client cacheing is optimal. E.g., if zou have configured your GW especially to send no-cacheing headers, then the all image, font, css, js files will be retrieved on every request. We found a header field being set by a WAS server to serve content with expire date of 01.01.1970 which caused the client to _always_ request _all_ static content over and over again.

Log performance

see Details for configuring that in Cognosw 11: https://www.ibm.com/support/knowledgecenter/SSEP7J_11.1.0/com.ibm.swg.ba.cognos.ag_manage.doc/c_ca_dgnst_log_overview.html

You can still use the ipfclientconfig approach in C11 - but see advice to not mix new C11 logging AND ipfclient approach. Attached find a config that especially logs the performance of APP + DB Server. Place the config in configuration directory, wait 30s to 60s and send a request to the server. Closeing the browser and reopeing a session to the server will eventually then write all logs to disk. All logfiles start with "AMVARA_xxxx". If the report runs two minutes and the UDA logfile has 1m 55s DB processing, then zou must have a look at your DB. from the logs you should be able to understand where the server is "loosing" time. Rename the file to ipfclientconfig.xxx to disable logging

Check BI process limits

The default settings for BI process is very tight and definitely limits execution of parallel reports. The max process in portal. I usually set them to (2x No. of Cores) -1

Check Dispatcher settings

There are settings that IBM Service / Support might give you like:

• RSVP.CONCURRENTQUERY.NUMHELPERSPERPROCESS=80 (default=0)
• RSVP.CONCURRENTQUERY.MAXNUMHELPERSPERREPORT=4 (default=1)
• RSVP.CONCURRENTQUERY.ENABLEDFORINTERACTIVEOUTPUT=true (default=false)

Depending on your issue configuring such a setting could solve your issue. I would not recommend setting these without knowing the details.

Do a loadtest

We have done a lot of load testing on C7, 8, 10 and 11 installations. Attached pls find a small testplan as startoff. It requests 2 reports cycling between them with parallel requests and rampup time to be configured. It runs against a PING secured Cognos environment. You most probably must adapt this to you needs.

Hope this helps.

------------------------------
Ralf Roeber
------------------------------

Attachment(s)

ipfclientconfig.xml   6 KB 1 version