[clue-tech] Downside of increasing TCP window size?

[Jim Ockers]

Hi Dave,

David Anselmi wrote:
> 
> Jim Ockers wrote:
> > Hi everyone,
> > 
> > We have a very low bandwidth, very high latency TCP/IP connection
> > that we're trying to optimize.  Is there a downside to using large
> > TCP windows on an Important web server, when only a few of the clients
> > are on high latency connections and most clients are on low latency
> > connections?
> 
> Can you set up a different server to mirror the site to the high latency 
> clients?  Then you could optimize for both.  I guess you only get one 
> set of knobs for TCP so you couldn't do this with different NICs or IPs 
> on your current server.

I am setting up a proxy server on the actual dialup terminal server.
That will essentially mirror the site for the high latency clients.

I mentioned the download is over SSL so I'm setting up Squid 3.0 which
will do HTTP to HTTPS conversion.  The clients would then be able to do
HTTP only to the proxy server, then the proxy server would do the HTTPS
to the actual server.  (We assume that the MSAT dialup connection is
fairly secure and we won't lose too much security/obscurity by using
HTTP only.)

We can tweak the kernel on the proxy server.  However we have to make a
software change to the remote clients to get them to use the proxy
server.  For clients that are "too remote" (i.e. it would be expensive
to change them) we are thinking about changing the actual web server's
TCP windowing as well.  The proxy server won't benefit those clients.

I think you're right that TCP tuning changes would affect the whole
system - there is only one set of knobs.

Our web application product manager is concerned that something will 
break for the terrestrial clients if we change the TCP window behavior
on the web servers.

Thanks,
Jim

-- 
Jim Ockers, P.Eng. (ockers at ockers.net)
Contact info: please see http://www.ockers.net/