Exchange 2010 has been optimized for inexpensive direct attached storage. The idea being that you replicate data among multiple inexpensive servers instead of introducing an expensive SAN. Many organizations are uncomfortable with this because they have bought into the idea that a SAN is much more reliable and faster than local disk ever could be.
However, SANs make me nervous because they become a single point of failure. It is not unheard of for a SAN to go down or experience performances issues. A couple of examples that students have given me over the last while:
- A SAN firmware update (installed by the vendor while the vendor is onsite) wiped an entire SAN. Restore of all the SAN data from backup took 3 days.
- A SAN with unreliable cache disks. This results in downtime while the cache disks are replaced and poor performance until they are replaced. This has been an ongoing issue for this SAN.
And, just for fun, disk in general can be influenced by vibration. Which makes sense when you think about it, but just your shouting can influence disk performance. Imagine construction vibration when piles are being set for a building next door.
No comments:
Post a Comment