So I've been on 4.1 now for a few weeks after languishing for too long on 3.5.
One of the biggest issues I was seeing to my environment on 3.5 was %RDY times floating anywhere from 2% to 8%, with an average around 3-5%. I was at a point where I really didn't want to consider many more 2 vCPU VMs or even just single CPU. I've vehemenetly turned down 4 vCPU.
Now under 4.1 something must be wrong. I rarely see 1% RDY times now. This is simply amazing...
Are you complaining?
If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
Indeed, VMware did a much better job with the CPU Scheduling in 4.x.
Interesting writeup on this in the "The CPU Scheduler in VMware ESX 4" white paper.
____________
blog.eeg3.net | Useful VMware-related Links
If you found this or any other post helpful, please consider the use of the Helpful/Correct buttons to award points.
Indeed - one of the biggest changes in 4 was significant improvements in coscheduler and relaxed coscheduling for multiple vCPUs.
The real kicker too was when they added multi-core for vCPU as well. Interesting read here.
http://www.boche.net/blog/index.php/2010/07/25/vsphere-4-1-multicore-virtual-cpus/
Cheers,
Chad King
VCP-410 | Server+
Twitter: http://twitter.com/cwjking
If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
Most definitely not complaining just sharing - I read all about the performance improvements and was itching for the longest time to get it done. I simply had no idea the change would be as huge as it was - for me. I mean think about it. I have roughly 33 VM/s per host, each host is 16 CPU (4x4 x 4 hosts) and a good spattering of 2 vCPU in the mix. I was regularly seeing the top 5-8 VMs on each host with %RDY in the range of 3-10% with some spikes for a couple of the 2 vCPUs to over 10% but nothing that was alarming to me. Now I rarely see anything over 1% across the board. Its as if the 'consolidation ratio' has been increased substantially. I know %RDY isn't the only factor to consider for how consolidated you go - but given limitless memory and storage, CPU availability is the only true constraint at that point and Ready Times define CPU availability.
But now the question about how many eggs do I want to put in each basket comes up. I feel like I could double the number of VMs per host and still not see the magic 5% number. But do I really want to put 60 VM's per host? Is management OK with that if the host were to actually go down?