HPE Storage Users Group

Hi All,

I have doubt in Adaptive Optimization.

As we all know the Adaptive Optimization is of 3 types.
1. Performance
2. Balanced
3. Cost

We implemented AO based on performance.

I want to know exactly how the movement of the LUN regions are done in this case.

Consider at some instant the hot region of the LUN is moved to SSD, will it stay in SSD itself or once it receives less IOPs will it move back to lower tier (FC/NL).

Please clarify my doubt as we are facing some issues in this implementation.

Regards,
KK

Data is moved based on access rate over time see whitepaper below, so if data becomes cold it may be demoted on the next scheduled move. The performance, balanced, cost settings just bias to the movement algorithm, you could also try adaptive flashcache to enable acceleration of data that's found itself in the wrong tier at the wrong time. But first have a look at the whitepaper to understand AO.

http://h20195.www2.hp.com/V2/GetPDF.asp ... 867ENW.pdf
http://h20195.www2.hp.com/V2/GetPDF.asp ... 731ENW.pdf

So performance based AO really just slows down the rate at which blocks are moved down. The fact is even in performance mode AO allows for a sustained 60ms access times on the NL tier which IMHO is way, way too high, and if HP would allow us to tweak the more fine grained aspects of AO I'd have set it to half that since my NL tier is like 80% utilized while I have an almost empty and underutilized Tier1.

You could try to optimize your settings. Check monitoring windows and region move times.
For best results you should only monitor times with high access rates and move those regions in times with low access rates. Do not monitor backup windows. With this in mind we made good experiences with AO (3 tiers).

best regards
Sebastian

Hi afidel,

be happy.
We have NL 100% Full and FC only 50%.
AO is set on Performance Mode and only watching Business Times (Mo-Fr).

For the System it is right, because all the Chunks on the NL Tier only have 0,01 IO Rate Density.