Let me take a stab at this.
Current symmetrix has 16 channel directors (kind of hbas on the storage side), each has 2 fibre channel ports. When you connect a server to this channel director, you use both for redundancy, so you can connect upto 16 servers. However, if you are trying to consolidate storage, then many more servers are likely to want to connect to this storage box. I don't know the average size of a SAN today, however, if I were to guess, I would say around 80-100 ports would be average. So with spider, EMC is trying to kill the external (small switch) market. That's why the name spider as a pun on Brcd's Silkworm. (Did they ever think about people who are arachnophobic;).
Now EMC's spider is kludge and competitive response to these new generation of storage systems out there from startups like Yotta Yotta, 3ParData, Cereva etc. See the storage folks are doing what was done in servers a while back. When you have devices hanging off a bus, the bus becomes the bottleneck. So you put a switch on the bus and then all devices have full bandwidth access to whoever they are talking to on the bus. The bus I am talking about here is the bus which is inside of the storage system. The disks are the devices. EMC's internal bus is still a bus and not a switched bus. The startups on the other hand came from the server background and said hey what not switch the bus. It will give increased throughput to the disks. EMC's spider is a response to this, but the switch is not on the internal bus but on to the I/O bus i.e FC.
Hope I didn't confuse anybody more, but that is what spider is, as you know spider eats the worm. |
