Integrated memory controllers, impact of memory fragmentation

[toha]

Intel Nehalem architechture has moved memory controller into CPU like AMD Opteron has had many years now. Has anyone conducted a study of performance impact of memory fragmentation with this architechture?

My point is that in theory it should be slower to access memory area which is controller by other CPU than which is actually running vCPU, but is this really an issue?

[mreferre]

It's basically a NUMA design (as it is the Opteron architecture).

ESX has been NUMA aware since.... 2002 at the very least so I don't think this is going to be a big problem (as it is not for the Opteron).

Massimo.

[Ken_Cline]

Yep...that's why you need to balance your RAM across all the CPUs. The only time you will likely feel the impact is if you have a VM whose vRAM allocation exceeds the RAM that is local to a given pCPU - then there's nothing that VMware can do (other than to use memory management features like TPS to reduce the RAM footprint)...

Ken Cline

VMware vExpert 2009

VMware Communities User Moderator

Blogging at: http://KensVirtualReality.wordpress.com/