Re: [SQU] Performance Question

From: Awie <awie@dont-contact.us>
Date: Thu, 2 Nov 2000 10:31:07 +0800

Henrik,

I implement the TCP header compression before I use Squid. The reasons were:

1. Bandwidth cost is very high.
2. Our users data "payload" is very variable.

Of course, our remote link should apply the TCP header compression (they use
same router; CISCO). Fortunately, our remote link also used Squid in Solaris
environment.

It seems our system runs well. I configure Squid as transparent proxy with
Brian's and your assist.

Thx

Best Regards,

Awie
awie@eksadata.com
PT. EKSADATA INTISOLUSI
Phone : (62-361) 261514
Mobile1 : (62-82) 3610369
Mobile2 : (62-818) 346241
----- Original Message -----
From: "Henrik Nordstrom" <hno@hem.passagen.se>
To: "Awie" <awie@eksadata.com>
Cc: <squid-users@ircache.net>
Sent: Wednesday, November 01, 2000 1:03 AM
Subject: Re: [SQU] Performance Question

> Sort of a similar thing.
>
> By turning on TCP header compression you increase the effective
> utilization of the link by sending more data in less bits.
>
> By running proxies on machines with good TCP implementations, satellite
> (or other very high latency high bandwidth) users can increase the
> effective utilization of the link by having a larger TCP window to keep
> the data flowing and correctly tuned retransmission timers to not have
> unneeded retransmissions only due to the high latency.
>
> Then you also have the interesting topic of proxies applying transport
> transformations of the data. Such as compression. This can even further
> increase the effective utilization of the link. There is a experimental
> patch implementing such things in Squid (and yes, HTTP allows for it,
> and it is even covered by standards)
>
>
> What you need to keep in mind for all those approaches is that there
> must be a proxy on both sides of the link for most of this, just as
> there must be support for compressed TCP headers at both sides..
>
> --
> Henrik Nordstrom
>
>
> Awie wrote:
> >
> > Wow......I become jealous to hear Audie's download rate. I only use 128
Kbps
> > through VSAT that (maximum) run at 10 KB/s to download.
> >
> > Henrik, it is interesting to read your email that having proxies can
> > increase the utilization of the link.
> >
> > I configured my router to have TCP Header compression that increase its
CPU
> > 2%-3%. But I can save bandwidth for around 10% - 20%. (Sorry, it perhaps
has
> > no relation with Squid discussion)
> >
> > Please advise. Thx
> >
> > Best Regards,
> >
> > Awie
> > awie@eksadata.com
> > PT. EKSADATA INTISOLUSI
> > Phone : (62-361) 261514
> > Mobile1 : (62-82) 3610369
> > Mobile2 : (62-818) 346241
> > ----- Original Message -----
> > From: "Henrik Nordstrom" <hno@hem.passagen.se>
> > To: "Audie Pierre" <audiep@vescomamerica.com>
> > Cc: <squid-users@ircache.net>
> > Sent: Tuesday, October 31, 2000 5:38 AM
> > Subject: Re: [SQU] Performance Question
> >
> > > Audie Pierre wrote:
> > > >
> > > > Hi! All,
> > > > If I am downloading @ 80Kb/s without the SQUID proxy server. Does
> > > > anyone know what the download speed should be approximately with
SQUID.
> > > > What is acceptable and what is not?
> > >
> > > 80Kb/s. Acceptable variance (except on cache hits) is plus minus a
> > > non-noticeable fraction. The exception is certain extreme types of
links
> > > (i.e. satellites) where having proxies can increase the utilization of
> > > the link due to lack of capabilities in most client and server TCP
> > > implementations and some other odd bieffects..
> > >
> > > --
> > > Henrik Nordstrom
> > >
> > > --
> > > To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
> > >
>

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
Received on Wed Nov 01 2000 - 19:28:44 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:56:13 MST