[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: RHEL4 running on VMWare ESX 2.5.2
- From: Tom Sightler <ttsig tuxyturvy com>
- To: "Red Hat Enterprise Linux 4 (Nahant) Discussion List" <nahant-list redhat com>
- Subject: Re: RHEL4 running on VMWare ESX 2.5.2
- Date: Wed, 26 Oct 2005 08:53:28 -0400
> We recently started running 2x IBM X445 (dual xeon/16gb ram) machines
> on VMWare. I wanted to run a few virtual linux machines from there due
> to testing and other stuff ....
Please note that RHEL4 is not yet supported on ESX although there are
some hints for running 2.6 based kernels on ESX in the VMware knowledge
base, I'm assuming you followed them? The most important seems to be
decreasing the Hard Timer period because otherwise ESX can't get enough
timer ticks to the running linux kernel.
> I installed RHEL4 U1 on around 4 machines. The performance is
> awwwwfuuull! ;)
I don't know what awful means, can you give some examples with hard
numbers? Load average doesn't really mean much, and seems to be
artificially inflated when running in a VM because the system time goes
up. I think this is due to the dynamic allocation of resource the
VMware does. You could try setting the minimum CPU%, and increasing the
timeslice (priority) of that virtual machine.
> Has anyone any experience of running RHEL4 (or any 2.6 kernel based
> distro) on VMWare ESX? I cant seem to find anything about this problem
> on Google ....
I'm running RHEL4 U2 (with a manual recompile of the LSI Logic SCSI
module from U1) and am getting decent, although not great performance.
Does your application have lots of threads? I've noted better behaviour
when booting with "clock=pmtmr" rather than the default which is to use
the tsc.
Also, if you are attempting to use SMP virtual machines you might want
to boot with the "noht" flag otherwise it seems that the kernel in RHEL4
detects the two "virtual" cpu's as a single, ht cpu.
With all of those tweaks I'm actually able to get pretty good
performance out of a VM running RHEL4. To give you some idea of what
I'm getting here is a very quick iozone benchmark from RHEL4 U2 running
in an ESX 2.5.2 VM (1GB allocated RAM) on a IBM x336 (Dual 3.2Ghz, 4GB
RAM) talking to an EMC CX400 storage array over the SAN:
KB reclen write rewrite read reread
2097152 8 99222 90961 66046 68286
So that's a 2GB file and I'm getting 90+MB/s write and ~66MB/s read, not
horrible based on the fairly old SAN hardware.
Now here are the same results from a similar physical server running
RHEL4 U2 talking to the very same SAN. It's a Dell 1850, Dual 3Ghz,
with 1GB RAM
KB reclen write rewrite read reread
2097152 8 88518 79269 46686 52762
Notice that my VM machine actually beat the physical machine. This is
probably explainable as the VM system is newer hardware, with faster
memory and processors, but it shows that I'm getting decent I/O out of
RHEL4 running on ESX.
Later,
Tom
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]