On the 29th of March 2010, HP released the latest version of SAN/iQ – 8.5
It’s a free upgrade to all 8.x customers and takes all but 10 minutes to upgrade a P4500 from 8.1 to 8.5 (including the node reboot)
8.5 most notably updates and enhances three areas of SAN/iQ:
1. Network RAID levels
2. Configuration and best practice alerting and notification
3. Application aware snapshots via a Windows VSS provider
In this post, I’ll have a quick look at the Network RAID changes
Fault tolerant volumes are the killer feature of SAN/iQ and the resiliency provided is its major attraction.
Prior to 8.5 you had the choice of Network RAID-10 with 2-way replica, 3-way replica or 4-way replica. i.e 2, 3 or 4 copies of your volumes distributed across the nodes for redundancy – the downside of this; decreased usable capacity.
8.5 introduces Network RAID-5 and RAID-6.
Network RAID-5 needs three data and one parity as a minimum configuration i.e 3+1 , meaning four nodes as a starting point.
Network RAID-6 needs four data and two parity as a minimum configuration i.e 4+2 , meaning six nodes are required initially.
Unlike Network RAID-10 which creates mirror replica(s) of a volume, the documentation states the new RAID levels stripe parity across all nodes in the cluster.
This in theory should mean less capacity utilization than Network RAID-10, but in my testing against a cluster of four VSA nodes it didn’t look that way.
A 5GB fully provisioned volume took 10GB of provisioned space no matter if created as Network RAID-10 or Network RAID-5 (see images). Edit – I note that in the docs it states “Network RAID–5 and and Network RAID–6 volumes require snapshots in order to achieve space utilization benefits” and it appears that snaps must be retained for at least a day to realize the storage savings or the new RAID levels. Taking a manual snapshot also has the same effect
Anyone else out there have the same experience with Network RAID-5?
Performance is still a question too. Until someone benchmarks Network RAID-5/6 on some hardware nodes a conclusion is hard to obtain.
Certainly, the latest HP P4000 SAN Solution user guide states that Network RAID-5/6 should be used for “mostly read, sequential
workloads on larger clusters, such as file shares and archiving”
This would seem to exclude it for VDI and shared VMFS datastore workloads as they are typically random 8K, with a large percentage of write activity in a lot of environments. Best to stick with the excellent performance of Network RAID-10 volumes
Certainly would like to hear from someone using these new RAID levels on a P4000 – please comment.
Edit: From the HP documentation:
Network RAID–5 and and Network RAID–6 volumes require snapshots in order to achieve space utilization benefits.
This means that deleting the last snapshot of a Network RAID–5 volume causes its space requirement to be the same as a Network RAID–10 (2-Way Mirror) volume.
Similarly deleting the last snapshot of a Network RAID–6 volume causes its space requirement to be the same as a Network RAID–10+1 (3-Way Mirror) volume.
It is possible, therefore, for the storage cluster not to have enough space to accommodate the snapshot deletion.
Deleting the last snapshot of a Network RAID-5 or Network RAID-6 volume is not recommended.