I'm trying to get my head around an element of the switch vs. director argument. It's claimed that a director can guarantee 100 Mbps between any attached host and its SAN based disk. If I had a Fan ratio of 20 FC connections between hosts and a director and 12 between disks and the director, is the claim that all 20 hosts can demand 100 Mbps simultaneously?
If so, please explain how this is possible (with the 12-20 ratio in mind), if not, please enlighten me. Also, my understanding is that a switch employs looping algorithms to handle heavy load traffic, is this the case and if so, how then does this traffic flow control differ from that in a standard hub?
Basically, any explanation on the actual flow of traffic through Directors, switches and hubs will be appreciated. Thanks in advance.
Part 2 of this question regarding fabric congestion:
For starters, I'm interested in your ratio of 20 hosts and 12 storage. You usually see ratios with more storage connections than host connections.
So, the question is how do you get full bandwidth with all this traffic through a single switch or director. The answer is a little bit complicated, but here are some variations:
1. You don't. It's a lie, a damn lie or, just a computer industry lie. 2. You do, it's not a lie at all, it's a great design and a superb implementation. 3. It depends on the application, the design and the implementation.
For starters, 100MB/sec is a theoretical maximum. Nobody gets that. HBAs and controllers don't communicate that fast. So cut it down to 75MB/sec and we have something closer to real transfer potential.
Then the question is how does all this traffic flow through a single switch? There are two answers, the backplane might actually be fast enough (and pigs fly over the Andes in April) or -- and this is where great design comes in -- a lot of the traffic doesn't have to go through a common backplane. How's that? The answer lies in multi-port chips that control 4, 8 or more switch ports. If traffic comes into one port and is destined to go out of another port controlled by the same chip, then the traffic doesn't have to flow over the backplane of the switch.
This can obviously have a huge effect on the overall transfer capabilities of the switch. The key is to set up hosts and storage on the switch so that "host-disk pairs" are connected through the same port-controller chip inside the switch. If the heaviest traffic moving through a switch never touches the backplane, the backplane doesn't need to have the aggregate performance of (75MB/sec*ports/2) where 75MB is realistic transfer rate and ports/2 represent the maximum number of end to end sessions through a switch at any time.
So, if the multi-port chip controller can handle the full bandwidth of all ports (75MB.sec*ports) on the chip and if the storage network can be established where most of the traffic is routed on the chip level - instead of over the backplane, it is possible that a switch or director could handle humongous amounts of traffic. The network configuration, including all redundant connections need to take this into account and should be a primary consideration in all network configurations.
Editor's note: For Part 1 of Marc's answer, go to http://www.searchStorage.com/ateQuestionNResponse/0,289625,sid5_cid400766_tax286191,00.html
This was first published in June 2001