Adaptive Optimization – let AO use up all SSDs or not?

[darienn]

Hi,

As titled, we use AO with SSD and FC 10K SAS tiers. Is there any reasoning not to let AO consume all of our SSDs by limiting its use of space? Or, would it be a better idea to let it use up all SSDs for performance?

BTW, below is our current configuration if it matters:

Two CPGs on our 3PAR with AO
1. RAID 6, with 14 data and 2 parity for 16 FC 10K SAS HDDs
2. RAID 6 with 4 data and 2 parity for 6 SSDs
- Default CPG: FC (moves to SSDs nightly with scheduled AO)
- Total raw capacity: 29TB
- Allocated space: >85%. (about 14% of allocated space is usused spare chunklets)

Thanks,

Darien

[Gary_L · Accepted solution]

Darien,

So with regard to allowing AO to use all the SSD space, if you are not creating any volumes in SSD and you are only using SSD for AO to move from FC then yes, you can allow it to use all the space, you do not need to set any limits.  If SSD is full, SO will just not move any additional regions to that tier.

Regarding your CPG configurations I would suggest reviewing a customer advisory that may pertain to your configuration,  Assuming you are running 3.3.1 and using express layout for both CPG's, the advisory would be pertinent.

https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-a00045368en_us

[darienn]

Thanks Gary for the advisory. Just one more thing to clarify before I click "Accept as Solution".

Does Express Layout also apply to the 10K FC SAS HDDs? I read that Express Layout is a feature associated with SSDs.

[Gary_L]

Express Layout was only available for SSD drives before 3.3.1.  As of 3.3.1 we opened up the solution to all drive types.

[Sheldon Smith]

Regarding Express Layout, it's easiest to see with the "showpd" command.

Here is an example of a system with FC drives where the system does not have them set for Express Layout:
Id CagePos Type RPM State Total Free A     B Capacity(GB)
0 0:0:0 FC 15 normal 278528 203776 1:0:1* 0:0:1 300
1 0:1:0 FC 15 normal 278528 206848 1:0:1 0:0:1* 300
2 0:2:0 FC 15 normal 278528 204800 1:0:1* 0:0:1 300
3 0:3:0 FC 15 normal 278528 207872 1:0:1 0:0:1* 300
4 0:4:0 FC 15 normal 278528 207872 1:0:1* 0:0:1 300
5 0:5:0 FC 15 normal 278528 211968 1:0:1 0:0:1* 300
6 0:6:0 FC 15 normal 278528 204800 1:0:1* 0:0:1 300
7 0:7:0 FC 15 normal 278528 207872 1:0:1 0:0:1* 300
8 0:8:0 FC 15 normal 278528 228352 1:0:1* 0:0:1 300
Note the asterisk (*) that shifts between the A and B Ports running down the list. That shows which controller node is considered the 'owner' of that physical disk and can own chunklets for logical disks.

Here is an example of a system with SSD drives enabled for Express Layout:
Id CagePos Type RPM State Total Free    A      B    Capacity(GB)
0 0:0:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
1 0:1:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
2 0:2:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
3 0:3:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
4 0:4:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
5 0:5:0 SSD 100 normal 1829888 1517568 0:1:1* 1:1:1* 1920
6 0:6:0 SSD 100 normal 1829888 1514496 0:1:1* 1:1:1* 1920
7 0:7:0 SSD 100 normal 1829888 1514496 0:1:1* 1:1:1* 1920
Note the asterisk (*) appears on both the A and B Ports. Either controller node can 'own' chunklets from that drive.