Sizing is actually pretty easy.  What Barton is getting at is that when
Linux reports performance numbers, it does so on the assumption that it
owns the hardware.  When running under VM (or LPAR), it doesn't so you
need something that understands VM's influence and can adjust the
numbers that Linux reports.  ESALPS can do this.

As to adjusting performance knobs, it's no different than any other
operating system.  Small, incremental adjustments until the natives stop
complaining.  If they always complain, that's another issue. :)

On Thu, 2004-06-17 at 12:58, Ranga Nathan wrote:
> Interesting. So it seems that if you want to track CPU and memory
> performance you have to do it at the aggregate VM level. That makes it
> difficult to size the guests. How does one know how much resource to give
> to a guest. I do understand that if the guest is under-performing then you
> give it some more. But by how much?
> Very intersting questions...!
> __________________________________________
> Ranga Nathan / CSG
> Systems Programmer - Specialist; Technical Services;
> BAX Global Inc. Irvine-California
> Tel: 714-442-7591   Fax: 714-442-2840
>
>
>
>
>
> Adam Thornton <[EMAIL PROTECTED]>
>
> Sent by: Linux on 390 Port <[EMAIL PROTECTED]>
> 06/17/2004 09:14 AM
> Please respond to Linux on 390 Port
>
>         To:     [EMAIL PROTECTED]
>         cc:
>         Subject:        Re: Performance Monitoring for Linux
>
>
> On Thu, 2004-06-17 at 10:29, Barton Robinson wrote:
> > I am unaware of any product other than ESALPS on the market
> > that is either aware of this problem or addresses
> > it.  Linux monitors do NOT have the ability to correct this
> > problem - they are unaware they are virtualized.
>
> They still have some utility: their absolute numbers--packets per
> second, for instance--are accurate.  And measurements of Linux processes
> vis-a-vis each other (i.e. "Samba appears to be eating about three times
> as much storage and four times as much CPU as named") are reasonably
> accurate, although that fluctuates a lot in any Linux system,
> virtualized or not.
>
> However, yes, as Barton says, anything that measures performance
> relative to the Linux image capacity is basically useless, since
> effectively the speed of the machine is varying wildly as VM does its
> thing (remember that what Linux considers core storage may actually be
> out in VM's page space somewhere, so not just CPU performance but memory
> access times are enormously variable).  So it tells you nothing that
> some process is consuming X% of the Linux image's CPU unless you're
> either running the image at a fixed priority--which I hope you're
> not--or you have some way to correlate that with what VM says the image
> itself is doing.
>
> Adam
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
> visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
>
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
--
Rich Smrcina
Illustro Systems International
rsmrcina at illustro.com

Catch the WAVV!
Update your zSeries skills in 4 days.
WAVV 2005 in Colorado Springs, CO
May 20-24, 2005
For details see http://www.wavv.org

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to