topic Question 6 node network raid 10 in StoreVirtual Storage

Question 6 node network raid 10

Ivanbre — Sun, 10 Oct 2010 15:54:42 GMT

Hi,

I got the following question:

Say i'm running a 6 node cluster of p4000.

I'm using raid 5 vor hardware level and network raid 10 for my volumes.

1st question:

Does it mean when i have network raid 10 my data only is written twice on two nodes? If so what happens if these 2 nodes in my 6 nodes cluster fail both at same time?

2nd question:

if i wont to expand my p4000 cluster. Do i have to have the same capacity as rest or doesn't it matter what capacity i purchase?

Re: Question 6 node network raid 10

Mark... — Mon, 11 Oct 2010 09:04:31 GMT

Hi,
1: 6 node cluster using N-10 then the same data blocks are written twice but it is written across all the nodes in the cluster:
1= AF
2= AB
3= CB
4= CD
5= ED
6= EF

2: If you add another node into your cluster then it should be the same or greater capacity. Can you add a smaller node? Well maybe depending on how much capacity has been used in the existing capacity.
For example:
3 nodes of 10TB = 30TB total
1 node of 10TB and 2 nodes of 20TB = 30TB total. The reason for this is that in a cluster the SAN/iQ s/w will default down to the size of the smallest node in a cluster. The rest of the space will be "lost" until all the nodes in the cluster are the same capacity.

Mark...

Re: Question 6 node network raid 10

Ivanbre — Mon, 11 Oct 2010 09:10:06 GMT

Hi Mark,

So what happens if node 1 and node 2 both fail.

I lose the written block A and A on both nodes? so i lose data right?

Re: Question 6 node network raid 10

teledata — Mon, 11 Oct 2010 10:53:35 GMT

True, in Network RAID 10 you could sustain data loss with the failure of 2 or more nodes.

As redundancy, and availability demands increase you can change the replication level (on a per volume basis).

You can go to Network RAID 10 +1 like this:

1= AD
2= AB
3= AB
4= CB
5= CD
6= CD

Now you can sustain a 3 node failure. You can go as high as Network RAID 10 +2.

Of course you are trading higher availability for storage usage/efficiency, but the nice option is that this choice can be made on a volume by volume basis..