How does a SAN avoid over-congestion?

[MadRat]

I don't understand how a SAN is magically supposed to avoid becoming congested on its network segment. What are people doing to prevent their SAN from becoming an also-run device on their local network segment? Seems to me that one or two users can easy overwhelm a SAN using multiple queries and applications no matter how one sets up the hardware. Are people using level 3 switches to balance loads? What else can be done? I'm not one to deploy SANs but the ones we have around here seem to be chronically slow. I'd like to hear the theories of how they ought to be optimized to prevent these g-awful slowdowns.

 

[Matthias99]

There's no magic involved -- SAN management is damn tricky. First off, the whole point of a SAN is that it provides alternate bandwidth just for storage needs (rather than your data being pushed back and forth across your LAN). However, as you have noted, it's still possible to have bottlenecks within the SAN itself if you're not careful with how it's laid out. Real enterprise-class SAN equipment can generally provide numerous multi-gigabit connections to hosts, which can themselves be switched onto, say, 10/100 ethernet, or simply more fibre links, in order to connect more machines to the network.

Systems that use a particular storage device heavily/continuously should be hooked directly to that device if possible. If the storage you're using doesn't offer multiple physical connections, you may need to look at a more flexible solution on that end, or, as you suggested, use load-balancing switches so that no one user can eat up all the bandwidth.