QUOTE(wildswing @ Nov 30 2006, 08:52 AM) [snapback]44815[/snapback]
Thanks for all the feedback fellas. It's much appreciated.
QUOTE(mcpaytas @ Nov 29 2006, 06:13 PM) [snapback]44750[/snapback]
You might think about ethernet ring switches. That way you get some fault tolerance.
Can you elborate for an ethernet idiot please? What's an ethernet ring switch.
There is a protocol called the "spanning tree protocol" and also the newer "rapid spanning tree protocol". These are IEEE standards that some switches have. There are also some proprietary versions of the same thing.
Normally, "loops" in your network are illegal. You have to wire everything as a tree. So this creates a situation where there is always a single point of failure everywhere...if a line goes down anywhere in the network, that portion of the tree is cut off.
A "ring" network eliminates that problem. Everything is in a single loop so if you lose any given link, it turns your ring back into a tree, but you don't have ANY downtime.
The spanning tree protocol effectively constantly maps your network's topology out automatically and allows you to do almost as many "loops" as you want. For instance, you could make a "star" network which allows for maximum bandwidth to and from the hub (preferred for SCADA) but connect the tips of the star together in a separate ring to have 3 alternative paths to and from any given point. A "ring only" protocol would totally freak out at this topology. A spanning tree doesn't care...it just takes the best route every time.
Another advantage of the spanning tree protocol switches is that those extra links can be something much less expensive and slower. For instance, make a star with fiber. Then the extra links could be wireless modems or CAT5E. Then if a fiber link goes down, you're "stuck" with a slow network around one of your switches but only one switch until you make repairs.
It all depends of course on how badly you want/need redundancy and how much you are willing to pay for it.