need explanation of Left Hand softwarel storage appliance and phisical P4000

[pob579]

Hi,

I am looking for the solution that will eliminate a single point of failure having place with one shared storage and server's fault tollerance allowing 1 node failure.

First, I want to ask about Left hand virtual storage appliance. If I understood correctly this software will allow to create a cluster from internal storages of 2 or more servers. HA and v-Motion (in case of vSpere) is available.

My main question is would it allow network Raid. If yes how the data will be distributed.

Please correct me if I get it wrong in my example below:

1. Let say I have 3 DL380 G7 running vSphere 5. Each node has 8x900GB SAS disks in Raid 10. That will give 3.6TB volume per server. Is it possible to create network RAID (kind of RAID 5), that will provide a fault tollerance (I can loose one server). And if yes, would the same formula of RAID 5 applied here.

So from 3 x 3.6TB, available space will be 2 x 3.6 = 7.2TB. Meaninig that I can use 7.2TB : 3 = 2.4TB on each server local storage.

2. What will be the best scenario to go with P4000 phisical storage for creating HA and if possible fault tollerance with minimal storage lost.

A storage need of the cluster for the next 3 years is 10 TB.

Thanks.

Michael.

[Bart_Heungens · Accepted solution]

Hi,

Regarding the functionality between the physical nodes and the VSA's it is exactly the same to the applied network RAID and so...

Regarding block distribution it is fairly simple. With the most commonly used network RAID 10, block A will be set on node 1 and 2, block B on node 2 and 3, block C on node 3 and 4, and block D on node 4 and 1 again (in a situation where U have 4 nodes).

In case of RAID5 it is the same, data blocks are spread accross the nodes together with a parity block...

In your 3-server setup, your 3 local disks will be seen as 1 SAN volume. If U go for RAID5, U will loose 1 time the size of the local disks of 1 server.

U can start with VSA now, U can add physical nodes afterwards in the same Lefthand cluster without downtime... Data will be restriped automatically...

Kr,

Bart

[oikjn]

i would add that there is little reason to use network raid5 unless the data is really low on the IO requirement.  Raid10 is fine.  It really is "One mirror" so with 3x 3TB stores, you should have something like 4.5TB of useable space and adding even a single additional node to a raid10 setup does increase the storage capacity.