More from Mike Laverick at RTFM Education on the last day of VMworld:
Performance on Citrix:
The actual session turned up no surprises from what we have all gathered from the Forum’s and experience. Single CPU VM’s for MFPS are the way to go; Turning off hyper-threading is still recommended; ESX 3.x offers much better user concurrency than ESX 2.x…. with ALL usual cavats that it depends on your APP and your user activity.
One of the outstanding session of my time here at VMworld. I really
couldn’t fault either content or the guy who delivered the stuff. He
really knew his stuff – and some more. I will be definitely taking some
of this stuff and re-working it into the classes I teach (time
When you get the chance to download the PDF of this session – make sure you do.
My only gripe – was some of my top questions, this guy wasn’t able
to answer. Like clunkiness of the iSCSI setup when you have CHAP
authentication; how you can’t reliably and effectively create vmkernel
switches with that X in the box to enable VMotion and so on. We were
treated to some “pat” answers on that which I just don’t buy.
Well, this is my 2nd or 3rd “troubleshooting” style sessions, and
unfortunately they are beginning to form a pattern. With very obivious
recommendations (such as document, be logical, read log files). This
one was a cut above the average session of this type. With a REAL
engineer talking about the REAL problems he has faced. Hey, what a
novel idea! Troubleshooting based on real experiences (plenty of irony
intended!) … So in short – for future reference. More examples. More war-stories. Keep it snappy…
This was a very interesting session. It took my already good knowledge
about some of the processor compatiablities to a new level. I would
strongly urge those who were unable to attend this years VMworld to
review the PDF of PPT when you have time. …
[Y]ou will be glad to know that I have taken the decision to try and
galvanise some of the Forum members who have been intelligentally
experimenting with this feature. Because of the hardware involved this
can’t be a one-man-rtfm-effort. It’s going to take a community
response. If VMware won’t or can’t give us all info we need – we’re
going to figure out for ourselves.