COGNOiSe.com - The IBM Cognos Community

IBM Cognos 8 Platform => COGNOS 8 => COGNOS 8 Administration => Topic started by: Cognos8 on 23 Mar 2011 03:43:01 AM

Title: Architecture suggestion
Post by: Cognos8 on 23 Mar 2011 03:43:01 AM

Hi,

In our environment we have 24 dispatcher, 2 gateway servers, one primary cm , 1 stand by content manager.

sometimes  environment  goes in freeze status , so to resolve  this we are thinking in multiple  ways including architecture change

In that Splitting the content manager is good idea? like dispatcher .. out of 24 do we need to have atleast 4 CM ? will it help to
improve the performance?


If you have seen this kind of architecture or related to CM split documents please provide.


version : cognos 8.4
Title: Re: Architecture suggestion
Post by: smiley on 23 Mar 2011 04:14:57 AM
If the environment freezes, and may require some sort of restart to come back up again, you may suffer from an overloaded CM indeed.
As you can only have 1 active CM, it will not help to add more, as they are only for redundancy, and not load sharing.

You probably allready have the biggest java stack defined right? (large configuration allocating 1536 mb)
Use the c8\bin\utilities\cm\CMTools\cmtools.exe to look how the memory in the stack is allocated, and behaving. If saturated, setup a 64 bits CM.

Rule of thumb for sizing the CM is 1 cpu for every 6 report server cpu's.
How is that currently setup?
Title: Re: Architecture suggestion
Post by: sashafay on 24 Mar 2011 12:57:17 PM
What is your high-affinity set on all 24 dispatchers? Try to lower a number (if it's set greater than 2), which may help to balance and stabilize your dispatcher's farm. Keep those numbers the same across all of your dispatchers though.
Title: Re: Architecture suggestion
Post by: AussiePete2011 on 24 Mar 2011 05:20:32 PM
Hi there

Just to add to this
Points to consider
Has this freeze occurred on a regular basis?
Does the freeze occur at a specific time of day?

What are the load characteristics of the server during this period?

Do you have dedicated batch reporting servers meaning that these servers only handle scheduled jobs, report etc?
Do you have a peak period defined for your dispatchers?

How many of your Application tiers have been added to the Gateways?

If you can start answering these questions only then should you start considering making changes to your current architecture.
Cheers
Peter B