evacuate preformance tier

[johnnoakes]

Hi.

We have a 3 tier MSA 2052

Controler A

Preformance = 6 SSD RAID6

Standard = 12 SAS10k RAID10

Archive = 10 SAS NL RAID6

We need more SSD so purchaching 4 more.

Q: Will RAID5 delever better preformance? If Yes, can I evacuate (remove disk groups) preformance tier whithout loosing data (provided there is space enough in standard and archive) And build a new 9 disk (power of 2) prefomance tier (with one spare). Or should I just expand the raid6 with 4 more ssd disks?

/Patrik

[JonPaul]

@johnnoakes 
It sounds like you are trying to get more performance overall out of your system. There are a few more knobs and switches to be worried about.  
First thing I would look at is the 'I/O Workload' graph in the MSA 2050 web UI.  This will give an approximation of the 'width' of your data usage and if it is inside or outside your SSD capacity.  
Next look at where pages are being allocated using the CLI> show tier-statistics
Third to look at is are all volumes the same importance of data accesse?  If not, can you set the volume tier-affinity to help prioritize (or de-prioritize) the volume?
As to your other question about DRAIN of a pool component.  Yes you can DRAIN any pool component as long as there is enough capacity in the other components to retain the data.  The caution here is that DRAIN is not a high priority for the system and can take some time and this is impossible to determine as it has dependecies on usage, capacity and other system tasks.  Once the DRAIN starts you can't stop it.  Also make sure that the DRAIN has plenty of margin capacity so that new data does not utilize the capacity the DRAIN was intending to take.  
Expansion of a disk-group is not possible,  expansion of a Pool is available by adding a new disk-group.

[johnnoakes]

Thank you.
What do you think of raid 5 vs 6 in performance tier?

[johnnoakes]

@JonPaul 

Also. The performance tier is 6x800gb (so 3.2tb) how long do you think a vdrain will take? finished over a weekend when there is also low iops?

[JonPaul]

@johnnoakes 
It's really hard to say.  It depends on a lot of factors and surprisingly I/O can actually help.
First - DRAINing disk-group performance, which is SSD so no problem there
Second - The receiving disk-group performance,  which could be the 10k and would be ok, or it could be the 7.2k which would affect overall performance
Third - The amount of data being Drained.  With a performance tier it is likely 90% or more full so you are going to have 3.2TB to move
Fourth - System load.
The DRAIN feature uses the tiering engine to do the migrations which helps it to not interfere with standard I/O, except tin this case the DRAINing disk-group will not be available for improving performance this will be significant with an SSD tier.  Looking at the Virtual Storage Technical Reference Guide it explains that the tiering engine will move up to 20 pages, a page is 4MB, every 5 seconds.  Which I did the math a long time ago, it turns out to be ~1TB per day.  So in that sense it could take 3 days to complete BUT if any of the data in the DRAINing disk-group is modified, then the re-written page will be put on a different disk-group which will help with the migration.  There is no way for you to tell what data is where so you can't strategically modify data.
There is really no way to tell how it is going to go until you are past the point of no return.  Once it starts you can look at the performance statistics, CLI> show disk-group-statistics   and see the number of pages de-allocated. And also look at the FREE space coming back in the disk-groups