Can we estimate consolidation ratios on new hosts?

[murphyslaw1978b]

I have two HA/DRS clusters with about 25 hosts and 300 VMs in each cluster.  These are running IBM HS22 blades with 96GB of RAM and have some 4-core CPUs, and some 6-core CPUs, mixed pretty evenly.

I recently bought and installed 6 new HX5 blades that are configured as wide blades that have 40 cores and 512GB of RAM on each blade. 

Comparing the two clusters, I'm wondering if I could easily "fit" all 300 VMs from one cluster into the new cluster.  Current CPU ready peak is bout 5%, and RAM is fine (not exceeding 75%).

Here's the comparison:

Current Cluster: 25 hosts running Xeon 5600-series, 768Ghz, 244 Processors, 2.34TB RAM

New Cluster: 6 hosts running E7-8800-series, 510Ghz, 240 Processors, 2.83 TB RAM

Also, on the new blades, I'm running 4 10-GB NICs and have 4 HBAs in round robin, so I'm not worries about throughput.  My thinking is that I should be just fine, since the total core/vCPU count is the same, I have plenty of RAM.  My only concern I guess is that the new blade have only 2.1Ghz vs. 2.9 or 3.3Ghz.  Most of my VMs are 2vCPU/4GB of RAM.  I have some that are 4 and 8 CPUs boxes.

[jklick]

If your concern is GHz, what is your CPU utilization like in your current setup? My guess is that like most environments, it's fairly low. If that is the case, then I would be more concerned with your vCPU:pCPU ratio (CPU Ready).

Where are you getting your CPU Ready measurement, btw? I'm yet to find an easy way to monitor CPU Ready in vCenter (per this blog post).

[murphyslaw1978b]

Jonathan,

Yes, my CPU utilization averages 15-20%.  My CPU Ready Peak is 5% at the moment.  I have Quest vKernel installed and it gives me reports on pretty much all performance metrics.  It just pulls stats from vCenter, and CPU Ready is there in vCenter, but it's only realtime on the standard statistic levels.  So vKernel reads these data and summarizes.   The only problem is that it's a 24-hour time period summation.  So I pick the heaviest days I can find.