[Xen-users] Hypervisor CPU load

[Andreas Pflug <pgadmin@xxxxxxxxxxxxxxxxx>]

Hi!

I wonder what is the suggested way to keep an eye on the hypervisor's
cpu load, to detect situations where more cpu cores should be thrown to
the given vm count, or vms should be moved to less loaded machines. All
information I found about performance monitoring seems to origin from
development or research sources, not meant for production use.

- AFAICS xentop only shows cpu usage, but won't give figures about vms
waiting for cpu time.

- xenmon together with xenbaked delivers the cpu wait time for each vm
slot using the trace infrastructure (if there are enough vm slots
compiled in... ). Apparently, xenbaked is called 100-200 times/second
and processing >1e6 records/second, so monitoring itself will probably
create quite some performance overhead.

- There's xenperf as well, requiring a recompiled hypervisor with
performance counters: I haven't checked if it will deliver the cpu wait
time or equivalent at all.

Did I miss something? Seems that xenbaked/xenmon style is the only way
at the moment?

Regards,
Andreas

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users