back to article Isilon bellyflops into Smart Pools

Scale-out file storage supplier Isilon is clustering its clustered systems into Smart Pools with automatic data movement between them to provide better balanced storage resources across storage tiers for its customers' applications. Isilon supplies scale-out clusters for file storage: the S-Series for IOPS-focussed …


This topic is closed for new posts.

Spellchecker strike

This paragraph seems to have suffered from an over-ambitious spellchecker:

Smart Pools have the different S-Series, X-Series and NL-Series nodes grouped or clustered together and linked by InfiniBand, Isilon's backed storage link…

Should read "Isilon's backend storage link"

It all sounds like just what the storage world has been waiting for, doesn't it.


What The Storage World Has Been Waiting For?

Hardly! So now when you want to add a new disk type or node type you can build a new cluster and link it to existing clusters....BFD! It's still only practical to scale out existing clusters with exactly the same node type with exactly the same type of disk. There are plenty of better solutions out there....and more cost effective!

Oh.. and isn't it IP over Infiniband on the backend storage link?


RTFM and it all becomes clear (IMHO)

Now when you want to add a new disk type or node type, you can add it to the _existing_ cluster so long as you add a minimum of three nodes of that type. You still have a single filesystem and you can still access any file from any node. If one pool of storage fills up, provided you have not unchecked the box, the data will spill over into another pool. It's up to you where you tell the cluster to put the files and when, and it's also up to you to make sure you have enough capacity in that pool if you don't want it to spill over. It is still one cluster, not many, and it is still managed as a single cluster from a single GUI (not click 'n launch). You cannot build a new cluster and link it because that would be two clusters and you can't join two Isilon clustered filesystems together.

Even if you're not using SmartPools for tiering, you can use it to improve your data protection. In a large cluster of 80 nodes, for example, you might like to set protection to n+3 (or more) as the likelihood of another node failing whilst one is down is is higher than in a small cluster of 20 nodes. With SmartPools you can split it up into four pools of 20 nodes and set the protection level to n+2 (per pool), which would use less space for protection and give a higher level of protection.

This topic is closed for new posts.


Biting the hand that feeds IT © 1998–2018