<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 9/22/2011 12:22 PM, <a class="moz-txt-link-abbreviated" href="mailto:scott.tanner@comcast.net">scott.tanner@comcast.net</a> wrote:
<blockquote
cite="mid:271076085.1782289.1316708565661.JavaMail.root@sz0161a.emeryville.ca.mail.comcast.net"
type="cite">
<style type="text/css">p { margin: 0; }</style>
<div style="font-family: Arial; font-size: 12pt; color: #000000">I'm
going to try using affinity to limit the CPU access and run a
few benchmarks - should be an interesting test.<br>
<br>
I've heard there's a fair amount of overhead when running MySQL
in virtual machines, specifically due to the abstraction layer
between the raid controller and guest. It would be interesting
to test, but difficult to setup now that the server is running
in production - It doesn't look like you can limit the number of
processors for Dom0.<br>
<br>
We purposely don't run anything else on these servers to reduce
any contention, so limiting the CPU's isn't a very big deal,
however wasting 4 / 16 cores (real / +HT) is pretty silly.<br>
<br>
</div>
</blockquote>
Have you tried just turning off hyperthreading? In quite a few
cases, the scheduler doesn't do a very good job of handling the
virtual CPU's of hyperthreading and actually makes things worse. At
best, HT is only going to help when you several different types of
tasks going on at the same time, like a floating point calculation
and an integer one. If mysql is mostly integer, which I suspect it
is, then at best HT is going to give you nothing. If the scheduler
gets it wrong and is trying to simultaneously run 8 operations on
only 4 real cores all you're going to do is double (or worse) the
amount of task switches.<br>
<br />--
<br />This message has been scanned for viruses and
<br />dangerous content by
<a href="http://www.mailscanner.info/"><b>MailScanner</b></a>, and is
<br />believed to be clean.
</body>
</html>