I love when the A-SIS functionality is brought up for NetApp.  The following is directly from NetApp’s Tech Report on Data De-duplication:  (http://media.netapp.com/documents/tr-3505.pdf)

1. “In its initial release, deduplication primarily had a focus on data retention/archiving of file system data
on secondary storage NetApp systems. Substantial storage savings can be achieved with
deduplication in some tier 2 primary storage environments as well.”  

What does that mean?  Not for PRODUCTION STORAGE VOLUMES.

2. “Is a background process that can be configured to run automatically, scheduled, or run
manually through the command-line interface.”

Post-processing data de-duplication is a thing of the past.  Most vendors who specialize in dedup are doing online dedup, why is this a concern?  It means that the overhead required for NetApp to do de-duplication was so high they couldn’t keep it as a running process.

3. “Is enabled on a per flexible volume basis.”

Ah, here is a serious limitation.  It is PER VOLUME ONLY.  Which means if you already have several volumes created in an existing FAS environment it will be much less effective.  Also, since NetApp’s architecture is clustered filer pairs, and most larger companies have several if not dozens of “pairs” each pair owns it’s own volumes, therefore you are stuck with much less de-duplication effectiveness.

4. .

So now we get into even better details.  If you are running the average size filers that NetApp continues to push into the Enterprise space you are very limited to Flex Volume size.  Why is this a problem?  Data de-duplication can only occur within a single volume.  So your expectation of running your entire VMWare environment on a single NFS share goes out the window, as well as all that supposed savings of disk space.  Oh by the way, these numbers are valid as of OnTap 7.4

5. NetApp licensing:

Deduplication is included in Data ONTAP and just needs to be licensed. Add the deduplication license using the following command: license add <a_sis>

If you want to run deduplication on any of the FAS platforms you will also need to add the nearstore_option license: license add <nearstore_option>

 

Deduplication is a licensed option behind the NearStore option license. Hence, in a clustered environment, both nodes must have the NearStore option and deduplication licensed.

Well so that is interesting, A-SIS isn’t just a license on its own, it is a sub-license of nearstore.  So watch out for those up-front licensing costs as well as on-going support costs.

6. Snapshot usage changes:

• Previous Snapshot copies will expire, and as they do some small savings will be realized, but they too will probably be pretty low.
• During this period of old Snapshot copies expiring, it is fair to assume new data is being created on the flexible volume and Snapshot copies being created.
• Thus the storage savings may stay rather flat (that is, very low).

I think this is one of the most powerful statements in the entire tech report.  YOU WILL LOSE ALL SNAPSHOTS when running the de-duplication process.  What does this mean for you?  It means forget about data recovery in a production environment if you want A-SIS running as well.

7.  Contradiction of CPU usages within the same white paper:

On Page 16:

Deduplication is tightly integrated with Data ONTAP and the WAFL file structure. Because of this, deduplication is performed with extreme efficiency. Complex hashing algorithms and look-up tables are not required. Instead, deduplication is able to leverage the internal characteristics of Data ONTAP to create and compare digital fingerprints, redirect data pointers, and free up redundant data areas, all with a minimal amount of performance impact.

On Page 18:

• If there is very little new data, run deduplication infrequently, because it doesn’t make sense to unnecessarily consume CPU resources. How often you run it will depend on the change rate of the data in the flexible volume.
• Stagger deduplication schedules for the flexible volumes so it runs on alternative
  days.

So which is it?  It is “extremely efficient”?  Or it is such an overhead hog that it needs to be treated with mittens, and carefully scheduled and not over used?

So, this is just information dissected from a public NetApp document.  There are plenty of other personal concerns I have over the A-SIS solution, but that wasn’t the point of this blog entry.