We have the following hardware voor our virtualization:
2x HPE Proliant DL380 Gen10 2x Cisco MDS9481s 1x MSA2052 (8x SSD 800 GB en 16x 600 GB SAS 15k), two controllers
The current config of MSA is: The MSA is connected via FC to the Cisco Fabric and then to HBA of Proliants There are 4x disk pools:
- Controller A: 4xSSD RAID10 en 8xSAS 15k RAID10 - Controller B: 4xSSD RAID10 en 8xSAS 15k RAID10 - There are 10 CSV's configured: 2x SSD and 3x SAS per Storage pool
QUESTION: - How to get the most performance out of the MSA? for our VM's and File servers
We have tested and we saw that the SAS pool was faster then the SSD pool. Is this normal?
- How about failover? - Should i balance the volumes on each controller? - Is there some Best practise for the CSV (Volume) size? We use for our SSD Volume 500 GB and for the SAS 750 GB
1st of all I want to correct few terminology otherwise it will create confusion for others who all are following your forum.
You have mentioned
"There are 4x disk pools:
- Controller A: 4xSSD RAID10 en 8xSAS 15k RAID10 - Controller B: 4xSSD RAID10 en 8xSAS 15k RAID10"
Here these are called Virtual Disk group (VDG) and not Pools
MSA2052 bydefault will have two Pools. One is for Controller A which is Pool A and another for Controller B which is Pool B
Above 4 VDG will be part of these two Pools. 1 SSD and 1 SAS VDG will be part of Pool A and another SSD and SAS VDG will be part of Pool B
You have mentioned
"- There are 10 CSV's configured: 2x SSD and 3x SAS per Storage pool"
You may have created 5 volumes per pool. 2 Volumes created from SSD VDG and 3 volumes created from SAS VDG per pool but actually at the time of volume creation it shows entire free space (Combine free space of all VDG) in that particular Pool. You just have to mention size of the volume. This is Virtualization so at the time of creation of Volume space will not get allocated as these are Thin provision volume. Space gets allocated as per demand created by hosts and write request.
MSA2052 having Automated Tier License already included so as per the usage of the volumes data will be moving from SAS to SSD and SSD to SAS tier automatically. So eventhough you think that you have created volumes from particular VDG but later 4MB pages will move between SSD and SAS tier based on data usage.
If you still want to define some preference or want to set Quality of Service (QOS) then you can define "Volume Tier Affinity" feature on each volume.
You can distribute 5 volume per Pool which means 5 volumes will be taken care by Controller A and another 5 volume will be taken care by Controller B. This is how load balancing will happen.
In case one Controller failed then another Controller will take care of all volumes.
You configuration is perfect and you just need little modification to get good performance.
You can also refer the best practices whitepaper to get more information on this,
