My colleagues are going to get tired of me saying this, but it is amazing to see in 2007 on small systems the exact same issues we faced in 1987 on our VM mainframes.

 

Of course, in 2007 we can acquire systems with massively large real memories.  And of course disks are much faster and more capacious, but anytime you start paging to disk you're slowing down dramatically.

 

One mainframe performance maverick back in the 80s said "If it is in your machine room and if it rotates, page to it."  Now you want sufficient real memory so you don't need your virtual memory to hit the disk at all.  Then you can hit the disk for other kinds of I/O.

 

(Of course I am citing hard-learned lessons from 20 years ago, with zero experience with modern VM systems....) 

 

/rich

 

On 7/30/07, Michael S. Surato <[log in to unmask]> wrote:

While this is not necessarily XEN dependant, here are a couple of our
observations:

1. Virtualization is memory dependant. On the systems we run, the
processors are idle most of the time, but the memory is almost always
full. As virtualization puts more systems on the physical hardware, the
constraints on the memory increase far more rapidly than the constraints
on other systems.

2. After memory, the bottleneck is I/O speed. In our systems I/O,
especially disk I/O, increases far faster than processor time. For this
reason, I would look to faster disk technologies, smarter RAID boards,
and higher disk caches.

This is not intended to downplay the need for more/better processors,
but simply to help prioritize the spending on the system being purchased
for virtualization.

+-------------------------------------------+
|            Michael Surato                 |
|      College of Arts and Letters          |
|      Michigan State University            |
|            320 Linton Hall                |
|        East Lansing, MI 48824             |
| Voice: (517) 353-0778 Fax: (517) 355-0159 |
+-------------------------------------------+
-----Original Message-----
From: MSU Network Administrators Group [mailto:[log in to unmask] ] On
Behalf Of Eric Weston
Sent: Monday, July 30, 2007 2:54 PM
To: [log in to unmask]
Subject: [MSUNAG] Red Hat Xen virtualization

Here's more questions for the virtualization gurus out there, in
particular those experienced with Red Hat Xen:

What is the optimal hardware for a server hosting multiple guest OSes?
are there certain processors that are preferable? Any to avoid? Is there
a rule of thumb regarding how much physical memory per guest OS? Any
other tips, warnings, advice, cautionary tales, caveats, monitions, etc?

   Thanks much,

                Eric W., Libraries Systems