After quite some time, I’ve been active @ Experts-Exchange and one of the questions that popped up was this;

have some concerns regarding network congestion.  Is there a formula out there for determining the necessary network capacity for X number of clients?

 

A very legitimate request if you ask me. When we have chips doing things for us, why should we tell the chips, how to efficiently do it? However, the nature of distributed solution is such that there is no central mechanism to govern it. Reminds me of British Colonies around the world – it just went out of their hands for the very reason :-)

As far as I know, there isn’t a formula which you can apply and resolves all these issues.

 

Basically, it is about good design till now, and as a matter of fact there are more and more research going on in this area;

http://www.freepatentsonline.com/5491801.html

The above is one patent on this idea.

Distributing the load, watching your network trend is the only way to say what mechanism is needed for your network to avoid network congestion. What I’m trying to say is, say you have a 500 node organization and I have a 500 node organization. There isn’t a single way where we can incorporate QoS or anything just like that -> It very much depends and only depends on your network traffic trend.

So different data points have to be taken. What is the average load of switches/routers/internet links and what kind of traffic is it? etc matters. Again there are different types of policies, we can have a fair share policies where each device gets a fair share, but in modern networks, we want a desktop browsing facebook to suck when it is in competition with a VoIP phone, isn’t it?

So start with your internet link utilization reports. Get MRTG (Or even better PRTG, 5 device license is free) to do an analysis to see how much you’re paying for your uplink and what your actual utilization is. Some times, the results can be shocking. Back in 2003, while I was doing contract job for a hospital, they had a twin T1 connection providing an effective bandwidth of 3 Mbps and their utilization for just about 20%. Still there were talks going on to get additional internet links for the reason that people complained about the browsing speed being slow!

When I entered the scene, I got caught in the momentum and was thinking of how to get better link. Then for another project, I was digging on the data set which revealed the utilization. Had to literally show to my VP on what is going on and that the internal network is the one which needs an operation (if not surgical :-) ).