WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-bugs

[Xen-bugs] [Bug 108] Xen domain scheduler allows only 50% physical cpu u

To: xen-bugs@xxxxxxxxxxxxxxxxxxx
Subject: [Xen-bugs] [Bug 108] Xen domain scheduler allows only 50% physical cpu usage
From: bugzilla-daemon@xxxxxxxxxxxxxxxxxxx
Date: Thu, 18 Aug 2005 15:47:07 +0000
Delivery-date: Thu, 18 Aug 2005 15:47:11 +0000
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-bugs-request@lists.xensource.com?subject=help>
List-id: Xen Bugzilla <xen-bugs.lists.xensource.com>
List-post: <mailto:xen-bugs@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-bugs>, <mailto:xen-bugs-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-bugs>, <mailto:xen-bugs-request@lists.xensource.com?subject=unsubscribe>
Reply-to: bugzilla-daemon@xxxxxxxxxxxxxxxxxxx
Sender: xen-bugs-bounces@xxxxxxxxxxxxxxxxxxx
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=108





------- Additional Comments From habanero@xxxxxxxxxx  2005-08-18 15:47 -------
OK, we seem to have a problem with time keeping on x86_64.  I noticed this when
most benchmarks on Xen get less than 1/2 the throughput than native Linux,
however all the reaim benchmarks did not.  The reaim benchmarks probably compute
time differently.  So, to make sure we had a trusted wall clock time reference,
I ran dbench but ssh'd for 'date' to another system before and after.  The
benchmark harness also reported local call to date:

               "ssh host date"                local date
              before    after                before    after
native Linux: 09:57:33  10:09:41 ~12 mins  00:00:13  00:12:18 ~12 mins
xen0 Linux:   10:18:14  10:24:22  ~6 mins  00:20:54  00:27:00  ~6 mins

So, wall clock timeis OK, but the xen0 dbench runs ran in 1/2 the expected 
time.  
Additionally, dbench reports elapsed time as it runs, and both runs looked like:

Running for 600 seconds with load '/autobench/sources/dbench3/client.txt'
  and ms16 clients started
  16      4761   513.24 MB/sec  warmup   1 sec
  16     11169   533.08 MB/sec  warmup   2 sec
  16     16500   534.30 MB/sec  warmup   3 sec
  16     21903   532.74 MB/sec  warmup   4 sec
  16     28053   532.70 MB/sec  warmup   5 sec
 <snip>
  16    671009   498.23 MB/sec  warmup 117 sec
  16    677205   498.55 MB/sec  warmup 118 sec
  16    683536   498.90 MB/sec  warmup 119 sec
  16    695893   554.03 MB/sec  execute   1 sec
  16    702155   542.27 MB/sec  execute   2 sec
  16    708169   536.73 MB/sec  execute   3 sec
<snip>
  16   4090403   492.85 MB/sec  execute 599 sec
  16   4096510   492.90 MB/sec  cleanup 600 sec
  16   4096510   492.08 MB/sec  cleanup 601 sec
  16   4096510   491.28 MB/sec  cleanup 602 sec

So, both dbench runs report running for 12 minutes total, but xen0 run really
ran for 6.  It sounds like domain virtual time is running at a rate 2x of wall
clock time.




-- 
Configure bugmail: 
http://bugzilla.xensource.com/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

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