Re[2]: [Xen-users] Re: Limit i/o capacitiy?

["Artem Y. Pervin" <ArtemPervin@xxxxxxxx>]

Hello,

Has anyone actually used the 'rate' parameter of vif?


> I`ve conducted a number of tests with the rate parameter of vif to
> investigate the effect this parameter does. I was expecting to see a
> proportional increase in performance of benchmark with the increasing
> of rate. But the experimental data shows different results.

> I`ve used a web-application as a workload and httperf as a benchmarking
> tool. There parameters of httperf were the following:
>> ./httperf --hog --client=0/1 --server=192.168.10.9 --port=50180 
>> --uri=index.cgi --rate=200 --send-buffer=4096 --recv-buffer=16384 
>> --num-conns=1000 --num-calls=1

> There key values of httperf's reports are:

> rate: 20KB/s
> test-duration 193.906 s
> Request rate: 0.0 req/s (0.0 ms/req)
> Net I/O: 0.0 KB/s 

> rate: 21KB/s
> test-duration 193.907 s
> Request rate: 0.0 req/s (0.0 ms/req)
> Net I/O: 0.0 KB/s 

> rate: 22KB/s
> test-duration 193.966 s
> Request rate: 4.9 req/s (204.0 ms/req)
> Net I/O: 15.8 KB/s (0.1*10^6 bps)

> rate: 25KB/s
> test-duration 193.966 s
> Request rate: 4.9 req/s (203.3 ms/req)
> Net I/O: 15.8 KB/s (0.1*10^6 bps)

> rate: 100KB/s
> test-duration 193.967 s
> Request rate: 4.9 req/s (202.7 ms/req)
> Net I/O: 15.9 KB/s 

> The performance (request rate) is 0 for rates 20 and 21 KB/s.
> The 22KB/s value gives normal results.
> However they don't improve much with subsequent increases of rate.

> Is this correct behavior? May be I have misunderstood something in the
> way the rate parameter should be used? Or passed wrong parameters to
> httperf?

> Thanks in advance.


> --
> Artem Pervin




-- 
С уважением,
Артем Первин
mailto:ArtemPervin@xxxxxxxx


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