Xen users - Mar 2007 - cpu life cycle comparison, virtualized to standard?

Hello to all,

Recently I got to thinking about how virtualization allows you to
maximize resource usage. 

On busy systems, this means processors and RAM are utilized to a higher
percentage all of the time. For 80% efficiency, one would need to be
utilizing at least 80% of all RAM and CPU on any given node 100% of the
time. This is easily achieved with something like Xen.

Have there been any studies that indicate how this type of utilization
effects processor life cycle? Do processors in servers that ran 2 years
not virtualized live longer than the same 2 in servers virtualized with
a high (80% or better) usage efficiency? 

Does anyone know if this is being studied and if data is becoming
available?

Best,
--Tim


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Tim Post wrote:
> Hello to all,
> 
> 
> Does anyone know if this is being studied and if data is becoming
> available?
> 
Google recently produced a report on disk failures.
http://storagemojo.com/?p=378
This doesn''t give an answer to your question though.

Purely on a personal level (ie not speaking for my company) the 
components which most often go wrong are the moving ones - disks and fans.
That aside, I don''t think running systems at 100% utilisation will 
reduce their lifetime.
THAT IS if cooling is adequate ie. if you make sure systems are cooled 
properly, proper front to back flow of cold air, and take steps to 
monitor the temperatures then I don''t think system lifetime will be 
reduced. Think of all the mainframes and supercomputers which are now 
either scrapped or sitting in storage - most of them probably still 
function fine, even though they had hard lives when they were in production.



John Hearns

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> -----Original Message-----
> From: xen-users-bounces@lists.xensource.com 
> [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of 
> John Hearns
> Sent: 14 March 2007 11:51
> To: tim.post@netkinetics.net; Xen-Users
> Subject: Re: [Xen-users] cpu life cycle comparison, 
> virtualized to standard?
> 
> Tim Post wrote:
> > Hello to all,
> > 
> > 
> > Does anyone know if this is being studied and if data is becoming
> > available?
> > 
> Google recently produced a report on disk failures.
> http://storagemojo.com/?p=378
> This doesn''t give an answer to your question though.
> 
> Purely on a personal level (ie not speaking for my company) the 
> components which most often go wrong are the moving ones - 
> disks and fans.
> That aside, I don''t think running systems at 100% utilisation will
> reduce their lifetime.
> THAT IS if cooling is adequate ie. if you make sure systems 
> are cooled 
> properly, proper front to back flow of cold air, and take steps to 
> monitor the temperatures then I don''t think system lifetime will
be
> reduced. Think of all the mainframes and supercomputers which are now 
> either scrapped or sitting in storage - most of them probably still 
> function fine, even though they had hard lives when they were 
> in production.
I completely agree. 

I have seen the "real" Mean Time To Failure analyzis done by AMD.
You''d
be surprised how long a CPU will last if it''s at 55''C. Problem
is that
there are other components that don''t last as long. Of course, using
the
CPU at higher than 55''C reduces the lifetime (quite noticably if
it''s
much above this temperature), but I wouldn''t worry about it still. It
will be VERY unfashionably ancient by the time it dies of old age. 

Memory, I expect, is similar. 

Of course, if the processor is OUTSIDE it''s valid operating
temperature,
all bets are off. But unless the cooling in your system is insufficient,
that shouldn''t happen. 

As a side-note, all of my systems are running BOINC-projects, which more
or less means that the system is running 100% CPU all the time. I have
not had a CPU-failure that wasn''t directly related to: 
1. Lightning strike
2. Heatsink failure/misuse (don''t expect the processor[1] to _not_
break
if heatsink isn''t attached). 

[1] Modern AMD processors have builtin overheating protection, but older
ones didn''t, as I found out the "empirical way".

--
Mats
> 
> 
> 
> John Hearns
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
> 
> 
> 


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users