Hi,
I am working on Cognos 10.2. My datasource is MSSQL 2012.My report performed about 1.5s in my testing. However, in production site, when 500 users run this report, it took 4.5 minutes to generate all requests. How can I boost performance of Cognos server ? Currently, my system included 6 CPU and 32G memory. :o
Thanks
Valiant
Quote from: valiant on 27 Aug 2014 06:01:12 AM
Hi,
I am working on Cognos 10.2. My datasource is MSSQL 2012.My report performed about 1.5s in my testing. However, in production site, when 500 users run this report, it took 4.5 minutes to generate all requests. How can I boost performance of Cognos server ? Currently, my system included 6 CPU and 32G memory. :o
Thanks
Valiant
You have 500 users running the same report concurrently? Really? What server sizing did you do to arrive at the spec you currently have?
Is it feasible to have the report scheduled to run once and save the output? That way the 500 users could simply view the saved output rather than running the report?
500 concurrent users is a lot! How many requests can your server handle concurrently?
MF.
Hi MFGF,
End-users need to enter a date from prompt page to run report. Each night, I already generated report in current year and last year. However, I want to test the case all users to generate report. Do we configure something to tune Cognos server in Cognos configuration or Cognos report service?
Thanks
Valiant
Quote from: valiant on 27 Aug 2014 06:44:20 AM
Hi MFGF,
End-users need to enter a date from prompt page to run report. Each night, I already generated report in current year and last year. However, I want to test the case all users to generate report. Do we configure something to tune Cognos server in Cognos configuration or Cognos report service?
Thanks
Valiant
Is it actually feasible that 500 of your users might want to run the same report at the exact same millisecond? How big is the entire user pool? Generally concurrency is calculated at a fraction of the total user audience since they will not all be running reports at the same time.
MF.
Does test have same data as prd?
Does test have same server/app/gateway configuration?
You can run the report in prd after hours when no one is running the report and see if it improves runtime for comparison.
If not, then something else is the issue.
May be playing around with high/low affinity connections will help.
Also, you can't just change CPU's without checking your server license agreement.