[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-users] Re: What does xm top mean by the following:


  • To: "pv" <vishnubhatt@xxxxxxxxx>, "xen-users" <xen-users@xxxxxxxxxxxxxxxxxxx>
  • From: "Petersson, Mats" <Mats.Petersson@xxxxxxx>
  • Date: Wed, 5 Jul 2006 12:03:29 +0200
  • Delivery-date: Wed, 05 Jul 2006 03:04:27 -0700
  • List-id: Xen user discussion <xen-users.lists.xensource.com>
  • Thread-index: Acafvl4RTuDGqLVpTtiu+C4OE296PAAWvCqQ
  • Thread-topic: [Xen-users] Re: What does xm top mean by the following:

One CPU will be able to give 100%, so 200% would be the result of 2 cpu's [which can also be seen by you having 2 domains running - you can only have one domain per CPU at any given time].
 
A CPU in this context is either anything the OS considers a CPU, which means hyperthreaded "virtual" CPU's are counted, just like cores on a dual core CPU. If you have one socket with two cores, which have hyperthreading, the OS will see it as 4 CPU's. In my test-system, I've got two dual core processors, and Xen sees that as 4 CPU's [I work for AMD, we don't do HyperThreading - as we have shorter pipelines...].
 
If you take one server using 35% of one CPU, and move it to a Xen-system, that domain would use 35% + x% of Hypervisor time - because we're now administering the domain through Xen, which means a little bit of extra work [or in extreme cases, quite a bit of extra work - depends entirely on the domain].
 
And as I said earlier, the processor(s) may well work 100% of the time, but the total in Xen-top isn't quite adding up to 100% * cpu_count, because some time is "lost". I don't know this for sure, but I've seen it before in other systems - some time isn't counted anywhere, although work is done...
--
Mats


From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of pv
Sent: 05 July 2006 00:04
To: xen-users
Subject: [Xen-users] Re: What does xm top mean by the following:

Thx Mats - I'm not quite sure what the %-ages mean i.e. what is 100% - is it the sum of all the cpu(hw or hyperthreaded or vcpus) on all the domains?
 
Take for .e.g a data center utilizing 35% of CPU for a particular server;
And, enter Xen: and you have installed 2 instead of one server running on the same h/w w/ the hope of increasing utilization. 
So, instead of 35%, one is expected to see 35%+x afa utilization (is this true?)
 
It does not matter what base-line or metrics one is presented with (I'm happy w/ the top presentation); as long as it is clear and easy to explain to a 'CIO, CFO or VP-finance or an unsuspecting fellow-worker' - if you know what I mean, I'll look forward to some documentation or an explanation to this regard. Thx in advance.
--

 
On 7/1/06, pv <vishnubhatt@xxxxxxxxx> wrote:
I'm compiling some tarballs in both the guest domains below (fedora1 and fedora2), they're quite compute/cpu intensive, when I do a xm top, I get the following, can someone tell me what does it mean to have CPU at 98.6% on one guest domain while 84.9% in the other when the host/domain-0 is at 3.2%? Thx in advance.
 
--
xentop - 20:38:02   Xen 3.0-unstable
3 domains: 2 running, 0 blocked, 0 paused, 0 crashed, 0 dying, 0 shutdown
Mem: 1038488k total, 1028408k used, 10080k free    CPUs: 2 @ 3391MHz
      NAME  STATE   CPU(sec) CPU(%)     MEM(k) MEM(%)  MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) SSID
  Domain-0 -----r         41    3.2     131100   12.6   no limit       n/a     2    8      879      378    0
   fedora1 ------         91   98.6     437652   42.1     442368      42.6     1    2        8       34    0
   fedora2 -----r         60   84.9     437564   42.1     442368      42.6     1    2        7       36    0
 

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

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.