Hi,
>> I notice that the xentop and xenmon tools do not try to combine the
>> CPU usage of an unpriveliged domain with the CPU usage done by dom0
>> on its behalf for IO. Has any work been done for determining the
>> "total" CPU usage of each unpriveliged domain?
I don't know, but it also would be very interesting for me!
>> For example, I may have an application running on a VM which is very
>> IO intensive. As a result, its CPU allocation may be very low, but
>> dom0 will be doing quite a bit of work on its behalf.
>> Presumably this could be estimated to some degree by looking at the
>> number of IO exchanges as reported by xenmon, but it seems like there
>> should be a more precise way.
Do you mean this value "0.02/ex I/O Count" ?
If so, what is the range of this value, I made some tests with dd,
netio, stress and others and got values between 0/ex and 11/ex but I
don't have any idea what is little, what is much and what is too much
(like 99% CPU load). How can I generate correct load for IO?
> I can tell you that this problem is being looked at. Some preliminary
> work was done in [1] which used I/O exchanges and there have been some
> follow up efforts since then. IMHO, there's no efficient way of
I've read this article, they explained how they measure it with Xen 2.0
but there is no code available and even if so it would not work with Xen
3.0.
So, is it possible to use the output of xenmon.py to monitor the IO-load
of a system for example to use it as a condition for loadbalancing? I'm
using already some output of xenmon.py for measuring the CPU-load, so it
would be nice if I could add also the IO-load to it!
Thanks a lot for your answers,
--
Chau y hasta luego,
Thorolf
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home