Planning Sever Scalability

[Amritha]

Dear Folks,

Can anyone let me know server scalability for the below:

1) Analyst Work Space ( How to consider?)
2) Application Sever CPU and RAM ?
3) Job Servers CPU and RAM

We are using windows 2008 OS and Cognos Planning 8.4

Please share the recommendations and Best practices.

Thanks a lot

A

[ericlfg]

Hi Amritha,

I'll attempt to answer your questions, although I'm not entirely sure I understand your questions specifically:

1) Analyst Work Space ( How to consider?)
This is the amount of RAM on the local machine reserved for the Analyst process.  As a best practice you typically do not want to exceed 1024MB as this would likely cause errors.  I usually set this value to around 512MB to start, sometimes increasing or decreasing this value may be required.  No two clients would typically have this setting the same.  This value will usually depend on the size of your d-cubes and other object aspects.

2) Application Sever CPU and RAM ?
If this machine will not be processing any jobs, then a lower powered server can be utilized.  A 2 CPU, 4 GB machine would usually be sufficient.

3) Job Servers CPU and RAM
If the machines are VM then please take a look at this proven practice document:
http://www.ibm.com/developerworks/data/library/cognos/infrastructure/operating_systems_and_hardware/page526.html

To summarize, it details that a larger number of smaller servers (4 cpus) vs larger servers (4 cpus+) perform better using a specific type of Intel processor.  In general the more CPUs you have available in an environment to process planning jobs will increase the performance and reduce the time required for completion.  Typically a general statement of 2GB per CPU is required.

Please let me know if I've answered your questions..

[StuartS]

Amritha

An issue we encountered with 8.4 was that an analyst workspace setting can sometimes be set too large.  As a Cognos Support Analyst put it, analyst will try to use a continuous block of RAM set by the workspace setting, make it too large and the machine will not be able to reserve a continuous block.  The ideal maximum we were advised was 768000.

Exmaple, a large analyst macro failed when the workspace setting was 1000000 but worked when it was 512000.

Stuart

[Amritha]

Thanks a lot Stuart.


Best regards,
A

[ovo]

Key rule is to scale RAM at 2GB per CPU core for any Planning Contribtutor Server.. and yes scale out rather than scale up.  More smaller machines is best, which is why VMWare (vSphere 4 onwards) is so useful.