Network Management

I'm being held hostage by the virtulization group. We have 20 machines on VM monitoring a total of about 30,000 devices. The environment is split 10 and 10 with each having it's own Master console and Failover.

When we were planning this system we were looking at deploying each machine with about 48GB memory, 10 CPUs, and 584GB storage. What got deployed is 8GB memory, 2 CPUs, and 160GB storage with the promise of increasing as needed, no choice in the matter but ok.

As soon as the machines were deployed and we started adding devices, about 2200/machine, we were starting to see performance problems, memory right of the bat then cpu. We were using Airwave performance charts and CentOS commands (free -m) to justify additional RAM but virtulization kept telling us their reporting showed no justification for the increases. In general vCenter memory utilization showed 50% utilization (no swap), and Airwave and CentOS showed 100% utilization and well into swap, 75% - 100%

So the question is why the discrepancy between the two and how do I justify use of my metrics?

On a side note, anyone using Cati RRDTools on these platforms?

I question the accuracy of the vCenter presentation of memory utiliation, and certainly the swap (as that would require the VMTools to be reporting that up and Linux's VMTools is notoriously 'spotty'). However, regardless, the 'Performance' numbers on the AMP > System page are showing you what the Machine is seeing, which as is the case is showing you it's starving for more power and RAM. 

Seeing how they gave you less than 25% of the necessary resources, I think it's clear they need to up the allocation. Or, you know, tell the PMO to just buy the damned hardware and then you guys won't ever have to deal with the vCenter guys again.

:P

PS I have a request into PLM to see if they have also seen this mis-match before, but I'm bettings it's a VMTools thing. What version of vSphere are they on, 5.0 or 5.5?