[ale] forwarded message from Mail Delivery Subsystem

[Vernard Martin]

> My theory is that HTTPD server load is mainly caused by the number of
> processes forked on a machine to handle requests plus any 
> processing done by cgi scripts.  One could imagine an
> architecture of web servers where a master server acted mainly as a
> gateway and load balancing.

[lines deleted for brevity]

> Another theory for making a fast web server would be to modify the OS,
> so that the slaves 'spoof' TCP connections from the master.  The master
> would recieve a connection and select a slave server to handle the load.
> The master would then communicate with the selected slave server, pass
> the parameters necessary for the slave to 'spoof' as the master server,
> and build the reply connections.  This eliminates the need for the 
> master to forward the reply from the slave.  The slave servers all
> pretend to be the external interface on replies.    
> 
> 
> Let's go implement this with Linux and make a few million dollars.


My phd research area is in paralle and dsitributed systems. I am current
doing a lot of work on the use of clusters of networked off-the-shelf machines
to get supercomputing bang per buck peformance. In a situation such as
this, you can actually get a way with it. Not too sh

I believe the SGI folks are simply running Power Challenege Arrays which
are several multiprocessors machiens connected via a HIPPI type bus but
using shared memory on each machines. Basically take 4 16-processor
machines connected via ATM speed connections and you can get great
performance. Since you actuallly have 4 machiens, you got 4 real I/O ports
and you have 16 processors on each machine to handle the large number of
actually threads that you need.

I'm ready to make millions if you are :-)

V
-- 
Vernard Martin
(vernard.martin at cc.gatech.edu) http://www.cc.gatech.edu/people/home/vernard/
College of Computing, Georgia Institute of Technology, Atlanta, GA 30332