[httperf] Doubts about parameter rate

Arlitt, Martin martin.arlitt at hp.com
Wed Oct 28 06:41:31 PST 2009


Hi Ricardo

In this case httperf is only generating a request rate of 102 res/second.  This is likely the limit of your server.  You will need to use the -timeout option for httperf to sustain a rate significantly higher than what the server can support.

Thanks
Martin


________________________________
From: ricardo figueiredo [mailto:ricardoogrande at gmail.com]
Sent: Wednesday, October 28, 2009 7:28 AM
To: Arlitt, Martin
Subject: Re: [httperf] Doubts about parameter rate

Hi Martin,

Thank you for reply.
I have another question.

For example.

# httperf --server 192.168.2.30 --uri /1 --num-conns 10000 --rate 500

Output:
...
Request rate: 102.2 req/s (9.8 ms/req)
...

In this situation of overload, How does httperf apply this parameter rate ???
Does HTTPERF keep the constant parameter --rate (500 req/s) ??? Or DoesHTTPERF decrease to 102.2 req/s ???

Thank you
Ricardo


On Sun, Oct 25, 2009 at 9:35 PM, Arlitt, Martin <martin.arlitt at hp.com<mailto:martin.arlitt at hp.com>> wrote:
Hi Ricardo

One problem is your server can't handle this load.  You should probably start with a much slower load and slowly increase it, to find what load your server can support.

If you use the -timeout option, your test duration should shorten quite a bit.  If you test at a high load like this, most of the connections will likely timeout.

Martin



________________________________
From: ricardo figueiredo [mailto:ricardoogrande at gmail.com<mailto:ricardoogrande at gmail.com>]
Sent: Sunday, October 25, 2009 12:56 PM
To: Arlitt, Martin
Subject: Re: [httperf] Doubts about parameter rate

Hi Arlitt,

For example, I ran the follow comand:

# httperf --server 192.168.2.30 --uri /1 --num-conns 10000 --rate 5000

The result is:

Total: connections 10000 requests 10000 replies 8302 test-duration 97.858 s

Connection rate: 102.2 conn/s (9.8 ms/conn, <=9646 concurrent connections)
Connection time [ms]: min 342.5 avg 20629.7 max 96020.1 median 15335.5 stddev 19709.0
Connection time [ms]: connect 12708.7
Connection length [replies/conn]: 1.000

Request rate: 102.2 req/s (9.8 ms/req)
Request size [B]: 66.0

Reply rate [replies/s]: min 0.0 avg 87.7 max 376.4 stddev 118.7 (20 samples)
Reply time [ms]: response 2943.8 transfer 2800.8
Reply size [B]: header 175.0 content 78841.0 footer 1.0 (total 79017.0)
Reply status: 1xx=0 2xx=8289 3xx=0 4xx=0 5xx=13

CPU time [s]: user 6.50 system 91.02 (user 6.6% system 93.0% total 99.7%)
Net I/O: 6553.1 KB/s (53.7*10^6 bps)

Errors: total 1698 client-timo 0 socket-timo 0 connrefused 0 connreset 1698
Errors: fd-unavail 0 addrunavail 0 ftab-full 0 other 0

Based on this result, Can you help me identify what (client, or network, or server) would be the bottleneck ?

Thank you
Ricardo

On Sun, Oct 25, 2009 at 5:37 PM, Arlitt, Martin <martin.arlitt at hp.com<mailto:martin.arlitt at hp.com>> wrote:
Hi Ricardo

It is difficult to know what the specific problem is without seeing all of the httperf output, or knowing anything about your test environment.  It is possible that your server, your network or your client cannot handle the rate.  Httperf can certainly handle that rate if the test environment can.

Martin


________________________________
From: httperf-bounces at napali.hpl.hp.com<mailto:httperf-bounces at napali.hpl.hp.com> [mailto:httperf-bounces at napali.hpl.hp.com<mailto:httperf-bounces at napali.hpl.hp.com>] On Behalf Of ricardo figueiredo
Sent: Friday, October 23, 2009 11:34 AM
To: httperf at napali.hpl.hp.com<mailto:httperf at napali.hpl.hp.com>
Subject: [httperf] Doubts about parameter rate

Hi,

My doubt's about parameter rate (--rate).
When I ran the follow commnad:

httperf --server www --ur /test.php --num-conns 10000 --rate 1500

I obtain the follow result:

...
Request rate: 96.0 req/s (10.4 ms/req)
Request size [B]: 66.0
...

Why is request rate 96 req/s in the result??
In command line I defined 1500 req/s.
Doesn't httperf execute the rate of 1500req/s during test ???

--
Thank you

Ricardo



--
Muito Obrigado

Ricardo



--
Muito Obrigado

Ricardo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://napali.hpl.hp.com/pipermail/httperf/attachments/20091028/941632f8/attachment.htm


More information about the httperf mailing list