Deduplication in HP 3PAR InForm OS 3.2.1 MU1

Today we updated our HP 3PAR P10400 array from InForm OS version 3.1.3 MU1 to 3.2.1 MU1. The big change here is the introduction of Thin Deduplication. Currently it only supports virtual volumes that reside entirely on SSD flash drives (no AO allowed), but word from our account team is that other media types are on the road map.

One of the most interesting features is the ability to run an analysis and estimate the deduplication ratio of data currently on a virtual volume (VV). Not every data type will be dedupe friendly, so this saves you and your disks the headache and wear of converting them to a Thin Deduplicated Virtual Volume (TDVV) only to find out it doesn’t save you anything.

To run the analysis (or “dry run”), open the 3PAR CLI and run:

checkvv -dedup_dryrun <vv_name>

This will start a background task that you can monitor from the GUI or using the “showtask” command. While it would be nice to initiate this from the InForm Management Console (IMC) and see the results in some sort of pie or bar chart, it seems the initial release is black & white.

When I ran it against one of our Microsoft RemoteFX VDI datastores (part clones, part SCCM-imaged VMs), it took a little over an hour to process a 1TB SSD VV. The result looked like this:

2014-10-30 14:50:16 CDT Finished checkvv process
 ------(MB)------ (DedupRatio)
Id Name Usr Estimated Estimated
92 vdirfx_ssd_a 778036 -- --
 1 total 778036 297475 2.31

That’s the end of the “showtask -d <task_id>” command.

Obviously most industry folks will balk at a meager 2.31:1 deduplication ratio for VDI, but I’d withhold judgment until more testing is available. Our VDI VMs have been moved around a lot and do not conform to the typical VMware or even Hyper-V clone metric typically used when marketing 7:1, 9:1, or higher ratios. These are persistent and patched systems that likely result in much more unique data.


Another value of the “checkvv -dedup_dryrun” process is the ability to run it against VVs currently residing on non-SSD disks. True, you can only dedupe VVs that are 100% on SSD/flash, but you can see what’s worth putting up there while the data is still on FC or NL disks. I’m running that now against a 3TB VV on FC 15K disks with about 1.8TB of actual data in the VV (it’s thin), and it’s running at about 5GB per minute. Thus, at this rate, it will be about 6 hours before I will have the estimate in hand.

Update 1:

2014-10-30 22:56:37 CDT Finished checkvv process
 ------(MB)------- (DedupRatio)
Id Name Usr Estimated Estimated
60 svr_db_b 1858433 -- --
 1 total 1858433 1589307 1.10

As you can see, MSSQL data isn’t ideal for deduplication on 3PAR at this time. Still, it’s good data to know!


  1. Simon said:

    I’ve been checking the support download site for 2 days now to see when the upgrade gets available. Where did you get the upgrade from? Did you have HP to do it for you since you got it so fast?
    I have the release notes already as they have been available since the 29th.
    Can’t wait to get the upgrade!

    October 31, 2014
  2. Chris said:

    Hey Simon,

    I learned about 3.2.1 from a sales presentation (re: 7400 series flash/hybrid arrays) last week, and put in a request to the 3PAR upgrade team ([email protected]) on 10/22. They were quick to respond and put me on the schedule for yesterday. Conveniently, 3.2.1 MU1 released yesterday as well, so we were able to skip past P01 and get on MU1 instead.

    Shoot an email over to them and I’m sure they’ll help you out. It was a smooth online upgrade and completed in <2hrs from start to finish (2 node system, including the code upload to the SP).


    October 31, 2014

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.