Ahh ok. That clarifies it, and my understanding of what we are going to do
leaves 2 CP's defined to VM and only one virtual CP defined to the guest,
with VM working out the details.




             Adam Thornton
             <[EMAIL PROTECTED]
             mine.net>                                                  To
             Sent by: Linux on         [EMAIL PROTECTED]
             390 Port                                                   cc
             <[EMAIL PROTECTED]
             IST.EDU>                                              Subject
                                       Re: Performance with Multiple CPUs

             06/24/2004 10:25
             AM


             Please respond to
             Linux on 390 Port
             <[EMAIL PROTECTED]
                 IST.EDU>






On Thu, 2004-06-24 at 09:51, James Melin wrote:
> Yes, that is the scenario we are planning to implement. When we added the
> 2nd IFL, I noticed that Linux automatically detected the 2nd CP at boot,
so
> I expect it to figure out going back to 1 just fine.

As long as you're running an SMP kernel, yes, it will work with 1 or
many virtual processors OK.  The thing I'm not sure I'm communicating,
though, is that I'm primarily interested in 2 cases.

In both these cases, the underlying VM system will be running on both
IFLs.  The difference will be whether 1 or 2 engines are defined to the
Linux guest.  (And not dedicated--that is, you just have multiple CPU
statements in the directory, but you're letting CP do the work of
deciding which actual engine performs the work on CPU 0 or CPU 1 of the
guest at any time).

> If you have any benchmarks you'd like to have me gather, let me know. I
> figure that such information will be useful for us in any case, and would
> also be helpful to the community at large. I expect to have this
> information gathered within the next week. Probably early next week

Really, as long as it's a workload that saturates the CPU and the I/O is
consistent between runs (be careful about disk caching here) I think I'm
fine.  Basically, what I want to know is whether you see any benefits
from the increased parallelism of running multiple virtual CPUs, as long
as the number of virtual CPUs is less than or equal to the number of
real CPUs.

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

Reply via email to