[Qatar-pm] [List] Does the Qtel filter affect web speed?

Kosala Atapattu kosala.atapattu at gmail.com
Sun May 3 21:01:58 PDT 2009


Interesting finding NIgel :).

I thought this is common within ISPs to cache internet traffic. This
functionality of Transparent Proxying is possible even with Squid. Some time
back this was a cool method to reduce the tier 1 bandwidth usage (most of
the big guys were talking ab it). For most of the countries (unlike Korea
and Japan... etc.) where internet traffic is not local, this was a must,
including SL. Well most probably they use a commercial tool to do that :).

Nigel,

It would be interesting to trace the network path it takes over port 80. I'm
stuck with a Windows box, can you give a try with hping on port 80 to do a
path trace.

# hping3 -S -t 0 -z -u 1 [IP Address] -p 80

You can use CTRL+z to increase the TTL. Or send me the IP address of the
server you tested, so I can play with it once I go home

SInce we got in to the subject of caching... people who still haven't looked
at Varnish, I suggest have a look. It was quite interesting.

Kosala

On Sun, May 3, 2009 at 4:37 PM, Nigel Gourlay <ngourlay at gmail.com> wrote:

> I'm ambivalent about posting this mail, but the figures probably don't lie.
>
> The Qtel filter acts as both a censor and as a webcache. At a recent
> Qatar GNU Linux User Group meeting, I suggested it would be
> interesting to see how the filter affected download speeds.
>
> Tested using 1024kbps ADSL, contacting a US-located webserver bound to
> ports 80 and 92. Port 80 traffic is intercepted by Qtel's filter. Port
> 92 traffic is ignore by Qtel.
>
> Static requests are HTTP GET requests for zero-byte and 100KB binary
> files. The responses should be cacheable by a webcache.
>
> Dynamic requests are HTTP GET requests to randomly generated URLs. The
> responses should not be cacheable.
>
> ICMP ping
> round-trip min/avg/max/stddev = 277.523/280.222/286.745/1.863 ms
>
> Zero-byte static requests
> Port 80: min/avg/max/stddev 60.673/68.117/198.970/18.626 ms
> Port 92: min/avg/max/stddev 581.702/598.875/1497.643/90.853 ms
>
> 100KB static requests
> Port 80: min/avg/max/stddev 1179.798/1446.170/1791.680/113.822 ms
> Port 92: min/avg/max/stddev 3745.255/4205.562/6069.717/563.729 ms
>
> Zero-byte dynamic requests
> Port 80: min/avg/max/stddev 323.830/382.158/1627.435/173.142 ms
> Port 92: min/avg/max/stddev 584.414/626.839/1716.627/130.719 ms
>
> 100KB dynamic requests
> Port 80: min/avg/max/stddev 1882.814/1914.729/2000.469/28.852 ms
> Port 92: min/avg/max/stddev 2450.092/2491.407/2530.966/17.465 ms
>
>
> Qtel's filter improves speeds for both static and dynamic requests.
>
> --nigel
> _______________________________________________
> List mailing list
> List at qglug.org
> http://lists.qglug.org/mailman/listinfo/list
>



-- 
Kosala
--------------------------------------------
Disclaimer: Views expressed in this mail are my personal views and they
would not reflect views of the employer.
--------------------------------------------
blog.kosala.net
www.linux.lk/~kosala/ <http://www.linux.lk/%7Ekosala/>
www.kosala.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.pm.org/pipermail/qatar-pm/attachments/20090504/9ad7a263/attachment.html>


More information about the Qatar-pm mailing list