Additional follow-up for item 1)
I was planning on filling the EXP24S with 24 ESFS - 1.7TB 10K RPM SAS SFF-2 Disk Drive 4 k Block (IBM i) Lose 2 to parity, 1 to hot spare - 36 TB usable, ASP # 2, mostly cold data.
For the CEC, I was planning on 6 (or more) ESOP - 775GB SFF-3 4k SSD for IBM I , Lose 1 to parity, 1 to hot spare - 3.1 TB usable, this would be SYS ASP 1, containing load source, Q libs, development tools, 1 copy of development data.
All controlled by 2 - EJ0L PCIe3 12GB Cache RAID SAS Adapter
Is this config acceptable/recommended? The 24 HDD in the EXP24S would see minimal usage.
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Sue Baker
Sent: Monday, November 30, 2015 7:27 PM
Subject: Re: Power8 S824 - mixing/placement SSD/HDD - is 571 gb 15K largest HDD drive available
<PSteinmetz@xxxxxxxxxx> wrote on
Wed, 25 Nov 2015 18:57:16 GMT:
1) If planning to mix SSD/HDD on a P8 S824 R&D LPAR, and using both
the CEC and a EXP24S, with EJ0L PCIe3 12GB Cache RAID SAS Adapter,
are there any pros/cons for placement of the drives, CEC or EXP24S?
If you think you'll need to grow beyond 26 387GB SSDs or you want to use 775GB SSDs, you need to place these in EXP24S with EJ0L and then use the 18 CEC drive bays for HDDs leaving the 8 1.8" bays empty.
While it is supported to mix SSDs and HDDs on the same controller, I recommend against it due to the potential for SSD operations to get queued behind HDDs operations which can cause disk response time anomalies.
2) Is 571 gb 15K the largest 15 drive HDD drive available?
Currently, it is. (Confirmed by looking at HW sales manual at
Currently on a P7 8205, using 571 10K, and seeing a large
performance hit using 5913 PCIe2 1.8GB Cache RAID SAS Adapter
Tri-port 6Gb. I'm shying away from using the 10k drives. I
know there are larger 1.1 tb 10k drives available.
Or, if my most used data resides on SDD, perhaps the 10k HDD
would be acceptable?
This may be acceptable in your environment. Only performance
testing with your workload can tell you for certain. I know of
no modeling tools to estimate what the results may be.