Re: 2500urls/sec and 150 Mbits throughput

Senthil,

        I was addressing the first part of your e-mail, the one that
mentions "2500 urls per second, with a throughput of 150 Mbits/second"
condition. Sorry for not answering the urls/second question, which I
misread.

As Henrik already pointed out, there is no difference between request
rate and URL rate because each Polygraph request is for a single URL.
HTTP defines that relationship to be 1:1 because one request message
has exactly one Request-URI parameter.

Polygraph does report request (URL) rate and bandwidth consumption for
both client and server sides.

HTH,

Alex.

On Mon, 31 May 2004, senthil wrote:

> u mean to say that setting 135req/sec(in polymix-4.pg) will generate
> 2500 urls/second?? how do i convince my client that 135 req/second
> corresponds to 2500 urls/second. is there any statistics in the
> polymix report relating to this info??
>
> thanx for the response.
> cheers,
> Senthil
>
> Alex Rousskov wrote:
>
> >On Fri, 28 May 2004, senthil wrote:
> >
> >> I have to run an urgent test on the cache and generate a
> >>report for my customer requirement. He asks to generate a report
> >>requesting 2500 urls per second , with a throughput of 150
> >>Mbits/second. Im not sure how to configure the polymix bed for
> >>generating the required result.
> >>
> >>is the requests/sec we set in polymix-4.pg in any way related with
> >>the number of urls/second?? if not could any1 help me with this.
> >>
> >
> >In PolyMix, you specify peak request rate. That rate, naturally, will
> >correspond to some peak throughput (proportional to mean response
> >size). Looking at last cache-off reports, 2500/sec load approximately
> >corresponds to 135 req/sec.
> >
> >You can modify PolyMix-4 response sizes and/or request type
> >composition to get the desired throughput for a given load. This
> >modification will require some experimentation to get the resulting
> >throughput right, since the traffic model is not simple. Your results
> >will not be, strictly speaking, PolyMix-4 results then. If your
> >customer needs PolyMix-4 results, they should not ask for both load
> >and throughput levels.
> >
> >HTH,
> >
> >Alex.
> >
>
>
Received on Mon May 31 2004 - 09:02:43 MDT