Rules for CLM on vpars in an integrity superdome

[Robert Cook_6]

Are there any best practices for configuring CLM for best performance in a vpar environment running global work load manager? One of my vpars will have a min of 2 and a max of 6 cpu's. How does CLM work if I start using cpu's from another cell board?

Thanks,

Bob

[Sameer_Nirmal]

I don't see any best practices white paper or something released publicly by HP for CLM setting. Maybe because the optimal setting of CLM depends on the type of application and workload the OS is running. One should know the application/workload behaviour to set the optimal CLM setting.

However applications using more than one cell would get benefited with the use of CLM.

vpar A.04.x allows use of ILM and CLM at the same time. Thus it adds to have optimal settings for your application/workload.
Refer this link for more details of them on
page 160 onwards.

http://docs.hp.com/en/T1335-90057/T1335-90057.pdf

Well CLM as it says "Cell Local Memory"(Non-Interleaved) . Here memory is not "striped" across the cells thus gives lower latency and improved performance. So if you enabled CLM on each cell, the CPU's on each cell would get low latency localized memory access and thus improves performance.

Now with gWLM which is basically CPU resource centric, you define the appropriate policy ( maybe utilization)therein. Thus even if gWLM attepmts to allocate additional CPU resource to meet workload's target and the CLM is set on the cell, the CPU would be benefited with localised memory access giving better workload performance.

[Robert Cook_6]

Thanks. Had to jump on another project.