%CSTP goes above 300% during snapshot delete

Hi all,

I am working with this huge, problematic VM:

• 12 GB RAM
• 12 CPU
• 2 disks (600 GB & 900 GB)
• 2 VMware Paravirtual SCSI controllers (0:0 & 0:1)
• No VSS

The problem is that after a backup which takes 30-60 Minutes, "deleting" the snapshot can take 4 hours and the performance is abysmal during that whole process. The only metric in esxtop that is out-of-line is %CSTP, and it is way out of line!

While %CSTP remains a firm 0.00 when our batch runs normally, if a snapshot is being deleted during the batch, we can see values of over 300%!

I am not able to successfully argue in favor of fewer vCPUs because the batch runs all 12 vCPUs to near 100% when it is running.

Any ideas?