HPE Storage Users Group

Ok, so I have several luns I'm moving from a 3par 7400, to a HPE Primera. ( I know not 3par 8k or whatever) but this relates to both.
On the 7400 I didn't have dedup & compression on the Primera which runs 3parOS v4 I do. I'm running windows 2012r2 with dedup enabled on the drives (it's how I had them set on the 3par)
I plan to move the luns to the Primera, is it best to continue to have windows dedupe? or undedupe in windows and have the Primera array deduplicate the data (which it would also compress?)

I am guessing the let the array do it. As it would offload the processing to the array. And then compressing and deduping of the data would be handled. Compared to windows just deduping..
Wasn't sure if there were any articles on this, as I have yet to find any.

Hello and welcome to the forums.

I concur, let the array do it! The windows server will only dedupe data on that one windows server, no global dedupe bewteen windows servers. The Primera will dedupe all data stored on it. Also the Primera has specialized ASICs to assist with dedupe hashing.

One article that you may want to look at is the StoreMore guarantee offered with Primera. The guarantee does not apply when the data is already deduped/compressed/encrypted by the host.

https://h20195.www2.hpe.com/v2/Getdocum ... skiphtml=1

Just some thoughts.

Windows does to my knowledge dedupe on file level (no fixed block size, or at least block size same as file system) so in some cases it might be more effective for a single server.

Dedupe comes at some cost either way. If your server or array is sized based on exact needs, dedupe might be a bad idea.

Personally I believe that OS level dedupe in a virtual environment is a bad idea and that array level dedupe is the better idea in most cases where it can handle the load.

But YMMV and the best advise is to just test. 3PAR has (and I assume Primera does to) a dedupe estimator so I don't need to go all in to know the result.

As a side note, my personal experience is that compression on some data types are better than dedupe. I recently did some work on a Vmware environment where I have no knowledge of what is running on the datastores. Half of the volumes gave a lot better compression than dedupe and the biggest space saver in total was compression.