[uml-devel] UML Guests and CPU Performance.

[uml-devel] UML Guests and CPU Performance.

[Primero]

[-- Attachment #1: Type: text/plain, Size: 1417 bytes --]

Hi ML, it is a little bit of time i've started using UML and i'm still 
"playing" with it. Anyway there's a thing that let me confused.

It seems to me that performance of UML Guest machine is sensibly worst 
then UML Host ... even if there is only one UML guest and no other 
processes sucking resources.

My Setup is:
Athlon 1000
512 RAM
1 GB tmpfs /tmp
30GB 4200 eide

The guest is started with 256MB of RAM

I can post /proc/cpuinfo:

Host:
model name      : AMD Athlon(tm) Processor
cpu MHz         : 1000.502
cache size      : 64 KB
bogomips        : 1970.17

Guest:
processor       : 0
vendor_id       : User Mode Linux
model name      : UML
mode            : skas
host            : Linux Pancho 2.6.8.1-cko8 #2 Mon Oct 11 17:03:08 CEST 
2004 i686
bogomips        : 1081.34


Here are results of Kernel Time compiling with the same .config:

Host:
real    15m27.226s
user    13m49.185s
sys     1m12.144s

Guest:
real    29m27.603s
user    21m16.970s
sys     7m59.080s


The Bogomips of the Guest are the half of the Host and compiling time is 
almost the double.

I want to know if this is "Normal" or i have to look for some errors in 
my configuration.

Thx
Primero

-- 
---------------------------------------
#######################################

"Linux, the choice of a GNU generation"
     ==Micro$oft - just say NO== 

#######################################
---------------------------------------


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 256 bytes --]

[uml-devel] Re: [uml-user] UML Guests and CPU Performance.

[Nuutti Kotivuori]

primero@fastwebnet.it wrote:
> It seems to me that performance of UML Guest machine is sensibly
> worst then UML Host ... even if there is only one UML guest and no
> other processes sucking resources.

It is definitely worse.

> The Bogomips of the Guest are the half of the Host

Ignore bogomips. Pure CPU workload is really close to the host. You
can try it with any CPU tester application.

> and compiling time is almost the double.

We have generally measured compile time increase by a third or
more. It is affected by many things though, so YMMV. This is with the
skas patch, but without SYSEMU.

> I want to know if this is "Normal" or i have to look for some errors
> in my configuration.

Sounds just about right - for some not-so-independent results on UML
performance compared to other techniques, I found the Xen performance
results to give a rough guideline:

  http://www.cl.cam.ac.uk/Research/SRG/netos/xen/performance.html

Do note that the benchmarks they run are really bad on virtualization
techniques - normal user workload suffers far less.

-- Naked




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel