Wednesday, September 2, 2009

TA3438 – Performance Improvements of vSphere…

Or something like that. Was Top 10 blah blah blah in the syllabus, doesn’t matter, it was good. Richard McDougall was really good at showing the performance improvements of vSPhere 4 (actually of ESX Server 4).

He broke them down into a few different categories and the major improvements there:




  • Relaxed CoScheduling
  • Extended Fairness Support


  • 2nd Generation Hardware Assist
  • Large Page Support


  • 350,000 IOPS up from 100,000 IOPS
  • Reduced IO Overhead by 50%
  • Asynchronous IO forks offload IO to a different core
  • Paravirtualized SCSI driver

He also talked about DRS improvements.

He made a comment that not all hypervisors are the same, and that just leveraging the hardware improvements of the processors isn’t enough. There is more to virtualization, and the vmkernel provides much in terms of access, scheduling, and IO access.

I agree that all hypervisors aren’t the same, BUT all hypervisors now give you quality performance near native, for most things, isn’t that enough?

No comments: