Re: req_rate V.S reply rate

From: Alex Rousskov (rousskov@ircache.net)
Date: Fri Dec 24 1999 - 23:49:05 MST


8823559,

        Req_rate field in PGL tells how often polyclt should try to submit
requests. Req_rate object in lx output is how often polyclt was able to
submit requests. Rep_rate object in lx output and the corresponding column
on the console is how often polyclt received a reply from the cache or the
server.

        You are probably misconfiguring the simulation. A single PolyMix-2
robot can use only 4 connections by default and hence cannot produce 100
req/sec streams. I think the following e-mails answer your questions and
give more details:

http://polygraph.ircache.net/cgi-bin/mail2html.cgi?fnames=19991201/424
http://polygraph.ircache.net/cgi-bin/mail2html.cgi?fnames=19991201/418
http://polygraph.ircache.net/cgi-bin/mail2html.cgi?fnames=19991101/388

Alex.

On Sat, 25 Dec 1999, 8823559 wrote:

> I am confused about these two statistics. Below is my console log:
>
> launch_win = 1
> ==
> for req_rate = 600/sec
>
> 001.51| i-warm 99 1.60 2544 12.50 0 4
> 001.60| i-warm 110 2.20 1891 36.36 0 4
> 001.68| i-warm 122 2.40 1740 41.67 0 4
> 001.76| i-warm 131 1.80 1990 22.22 0 4
> 001.85| i-warm 143 2.40 1533 58.33 0 4
> 001.93| i-warm 156 2.59 1654 38.46 0 4
> 002.01| p-warm 168 1.40 1881 33.93 0
> 002.02| i-inc1 168 2.39 1692 33.33 0 4
> 002.10| i-inc1 178 2.00 2187 30.00 0 4
> 002.18| i-inc1 188 2.00 1732 40.00 0 4
> 002.27| i-inc1 203 2.99 1306 33.33 0 4
> 002.35| i-inc1 210 1.40 2724 28.57 0 4
>
> ==
> req_rate = 100/sec
>
> 000.60| i-warm 0 0.00 -1 -1.00 0 0
> 000.68| i-warm 0 0.00 -1 -1.00 0 4
> 000.76| i-warm 9 1.80 2013 22.22 0 4
> 000.85| i-warm 18 1.80 2021 33.33 0 4
> 000.85| fyi: server scan completed with all 1 local robots ready to hit
> all 1 ss
> 000.93| i-warm 31 2.59 1507 38.46 0 4
> 001.01| i-warm 45 2.79 1587 28.57 0 4
> 001.10| i-warm 55 2.00 1210 50.00 0 4
> 001.18| i-warm 65 2.00 3033 20.00 0 4
> 001.26| i-warm 70 1.00 2815 0.00 0 4
> 001.35| i-warm 81 2.20 1908 45.45 0 4
> 001.43| i-warm 100 3.79 1098 63.16 0 4
> 001.51| i-warm 109 1.80 2168 22.22 0 4
> 001.60| i-warm 119 2.00 2031 30.00 0 4
> ̣got shutdown signal (2)
> 001.67| noticed shutdown signal (2)
> 001.67| resource usage:
> CPU Usage: 13.930sec sys + 27.000sec user = 40.930sec
> Page faults with physical i/o: 285
>
> 001.67| fyi: current state (2) stored
> 001.67| got 127 xactions and 0 errors
> 001.67| shutdown reason: got shutdown signal
>
>
> ==
>
> Could you please tell me what's the relationship between req_rate
> and reply rate? I find that they are rather different from each other.
> (The reply rate is much lower than req_rate)
>
> The response time is rather high . I wonder if I can trust the
> resulting response time I get.
>
> Maybe the squid server don't has so much tolerance? (for req_rate
> higher than 100)



This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:10 MDT