New SAN design

Not really clear on what you are asking as the alternative option, but for the most part we still stick with the tried and true core-edge topology. It is predictable, adaptable, and growth can be easily managed over time as long as you plan for growth capability up front.

As with everything though, it all depends on the use case. You may have a situation where a full mesh is absolutely the right decision. Or it may make the most sense to go with larger director class switches and NO edge. Can't really tell what is best without lot's of detail on the current and future use case.

san design is always depends on budget, cost, efficiency, capex. performance, as Allen said, it's really hard to tell what is the best solution to fit your environment, there are some white papers at Brocade web site, also, if Cisco, you might want to check cisco live presentations, that would give you some idea.

to add ..

what is the current capability of environmnt, like speed of the switch port and speed of the HBA for the hosts connected?

we were with core-edge model where core provided 2Gbps connections while edge provided 1Gbps connctions; Now moved to 8Gbps capable switches( no more edges), but most of our critical servers are still running with 4Gbps HBA but they were either running 2Gbps or below before that.

other area is number of ports or scalability. Most of the director class switches comes with 4 different types of blades (based on my Brocade knowledge)

1.switch blade is the port blade where is insert the SFP

2.CP (control processor) blade is the one contained FOS

3.core blade is the core switch blade which is used for ICL( inter Chassis Link) interconneting 2 or more devices)

4.AP blade is the application processsor

you need to know how mancy SW blade your model can support; like you can with 6 SW blades(6*48 =288 ports) now if needed you can add additional SW blades..

what did u choos to go with finally?

We have not decided yet but a few guys want to go with an edge switch(4G) for every cabinet in the DC and then having 2 core's with 8G ports.

Not sure but I think this is a mesh topology.

The argument is that there is lot less cabling we need to take care of because all the hosts in that cabinet attach to the switch in it and there is ISL's already established when we set it up. More like peice mailing the cisco USC model.

If UCS is a great idea them I guess this would work fine also.

This sounds like a core-edge to me, but it all depends on where your ISLs are. I'm assuming the ISLs would be from the switches in each cabinet back to the core switches and not between switches in cabinets.

A true mesh would have every switch ISLed to every other switch (even server racks). They are highly resilient, but not very scalable).

Sounds like you are on the right track.The only concern I have is that you are talking about (or it sounds like you are talking about) a single edge switch for each rack. That doesn't provide much redundancy for the hosts even if you dual attach them. It kind of defeats the purpose of having two core switches since a single rack switch failure could take out that entire rack's connectivity.

If you deployed the two core switches (not connected to each other) and then two edge switches in each rack, each one connecting back to one of the core switches... you would end up with fully redundant connectivity for your hosts as long as each one connected to both of the edge switches in that rack.

If the port count on the switches is to high to justify that you may want to consider end-of row switches instead of in rack switches.

As I stated earlier though there are a LOT of factors that go into the right SAN fabric design for a given site. What is perfect for one site might be absolutely wrong at another site.