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-devel

[Xen-devel] Xen 2.0 performance numbers

I thought people might be interested in some benchmark results I
recently recorded.  I've appended the output of a run of the OS
micro-benchmark lmbench 3.0, comparing the performance of Linux
2.6.8.1 running native, and on Xen.

The figures show that the performance of Linux running on Xen is
very close to that of native.  On many benchmarks, the overhead
of running on Xen is actually less than the overhead of an SMP
build of Linux vs a uniprocessor one.  (Xen itself is SMP
capable, but each of the guest OSes must currently be
uniprocessor -- we'll fix this post 2.0).

Xen shows some overhead on the context switch, fork, and page
fault benchmarks, but that's the price of safe virtualisation.

Remember that these OS-intensive micro-benchmarks represent an
absolute worse case scenario for any kind of virtualization
technology.  Real-world workloads will typically show rather
lower overheads (within a few percent for IO-intensive
applications like Apache and PostgreSQL, and less than 1% for
computer intensive applications like Povray etc).

Anyhow, I'm off on holiday until the 19th, so please send email
to the list rather than direct to me.

Best,
Ian

=====================================================================

key:
 26-nat-sm : Linux 2.6.8.1 native SMP build
 26-nat-un : Linux 2.6.8.1 native uniprocessor build
 26-xR1301 : Linux 2.6.8.1/xen (uniprocessor) on Xen version 1.1301

These figures were recorded on a dual processor 2.4GHz Xeon box,
with SMP Xen.


                 L M B E N C H  3 . 0   S U M M A R Y
                 ------------------------------------
                 (Alpha software, do not distribute)


Processor, Processes - times in microseconds - smaller is better
------------------------------------------------------------------------------
Host                 OS  Mhz null null      open slct sig  sig  fork exec sh  
                             call  I/O stat clos TCP  inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ----
26-nat-sm Linux 2.6.8.1 2400 0.49 0.62 2.25 3.72 24.3 0.90 3.49 189. 891. 3043
26-nat-un Linux 2.6.8.1 2400 0.46 0.52 1.46 2.27 5.90 0.79 2.76 114. 590. 2454
26-xR1301 Linux 2.6.8.1 2400 0.46 0.52 1.29 1.98 5.44 0.79 1.34 218. 879. 3183


Context switching - times in microseconds - smaller is better
-------------------------------------------------------------------------
Host                 OS  2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
                         ctxsw  ctxsw  ctxsw ctxsw  ctxsw   ctxsw   ctxsw
--------- ------------- ------ ------ ------ ------ ------ ------- -------
26-nat-sm Linux 2.6.8.1 5.5300 5.5600 5.7700 5.4500   28.7 3.83000    31.7
26-nat-un Linux 2.6.8.1 1.3400 1.4900 1.6100 2.3900   25.5 5.78000    39.2
26-xR1301 Linux 2.6.8.1 3.1600 3.4900 3.6400 4.2600   30.3 8.92000    41.3


*Local* Communication latencies in microseconds - smaller is better
---------------------------------------------------------------------
Host                 OS 2p/0K  Pipe AF     UDP  RPC/   TCP  RPC/ TCP
                        ctxsw       UNIX         UDP         TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
26-nat-sm Linux 2.6.8.1 5.530  15.0 18.9  26.4  35.0  29.1  37.8  55.
26-nat-un Linux 2.6.8.1 1.340 5.449 8.79  12.0  19.9  12.7  24.2  42.
26-xR1301 Linux 2.6.8.1 3.320 8.993 16.9  18.2  26.4  19.2  30.3  53.


File & VM system latencies in microseconds - smaller is better
-------------------------------------------------------------------------------
Host                 OS   0K File      10K File     Mmap    Prot   Page   100fd
                        Create Delete Create Delete Latency Fault  Fault  selct
--------- ------------- ------ ------ ------ ------ ------- ----- ------- -----
26-nat-sm Linux 2.6.8.1   41.3   19.0   96.0   38.3   134.0 0.889 2.37650  22.4
26-nat-un Linux 2.6.8.1   32.2 6.1050   69.8   15.0    84.0 0.871 1.61550 3.614
26-xR1301 Linux 2.6.8.1   32.0 5.5340   64.9   14.3   155.0 1.372 2.67530 3.016


*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------------------------
Host                OS  Pipe AF    TCP  File   Mmap  Bcopy  Bcopy  Mem   Mem
                             UNIX      reread reread (libc) (hand) read write
--------- ------------- ---- ---- ---- ------ ------ ------ ------ ---- -----
26-nat-sm Linux 2.6.8.1 705. 477. 313. 1411.8 1550.0  572.0  577.6 1536 788.3
26-nat-un Linux 2.6.8.1 1031 1646 364. 1551.4 1550.3  568.5  574.3 1548 796.8
26-xR1301 Linux 2.6.8.1 526. 1079 323. 1510.1 1568.3  550.1  550.4 1566 786.4




-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/xen-devel

<Prev in Thread] Current Thread [Next in Thread>
  • [Xen-devel] Xen 2.0 performance numbers, Ian Pratt <=