Questions re Centera Collections

[Cheryle_Weather]

Hi All... I tried to locate info with this new format and didn't have much luck.  Support was not much help either.

We are currently running EV 7.5sp1 -- with plans to upgrade to sp5 soon.

Unfortunately, we have run into the object limitation issue on the Centera.  Rather than purchase new storage, we are looking at ways to downsize the number of objects and eventually upgrade the Centera to remove that object limit.  I am currently deleting old archives and vaulted items, but thought enabling collections would provide faster relief.

I currently have 2 Centera partitions... one of which is closed.  I assume to implement collections, I would need to close the current partition and open a new one with collections enabled?  If I turned collections on the (then) closed partitions, would there be anyway to back out of it?

Is there any noticable performance degredation for day to day operations with collections running (IIRC this is 24/7 since it cannot be scheduled with ev7.5).

Any assistance and/or suggestions would be most appreciated =)

Cheryle

[Joseph_Rodgers]

Cheryle,

Funny, we were just debating using collections at 2 different clients and the pro\cons for them.  I've found strong opinions on both sides but it sounds like you are out of options.

I believe (not 100% positive and cannot check right now) that collections can be enabled on the existing partition if you so desire.

The centera collection should have some dedicated disk (preferrably its own lun) and based on various best practice guides I would go with ~150GB for size.  Most likely this storage will be grossly underutilized but it is better to have it than not.

The performance impact might suprise you.  Archiving tasks may actually run slightly quicker as they will write to the  collection area first.  However restore\retrieval may be slightly slower (I doubt users would notice).  Additionally if you require DOD wipes or need to be 100% certain that the data has been destroyed upon deletion than you should not use collections.  Collected files that are "deleted" technically still exist in the collected object until the object itself is eligible for deletion.

You don't have to worry about scheduling collections for Centera, it works very differently than NTFS collections and will occur automatically whenever the archiving task is run.

If you enable collections on a new partition and then closed that partition that should be ok but once collections is enabled I don't think it can be disabled.

-Joe

[TonySterling · Accepted solution]

With Centera collections the files are collected as they are archived, not according to a schedule.  So enabling collections will not work retroactively but only collect newly archived items.

I do however agree that using collections is a best practice and have verified with EMC that fact.  

[Cheryle_Weather]

Hey Tony!

So, if I'm understanding correctly.  We could not "collect" what is already stored in the Centera partitions.  If we enabled it, collections would occur on a going forward basis only?

Thanks!

[TonySterling]

Good to hear from you!

Right, items already on the Centera will remain untouched but moving forward your mail would be collected.  I recently chatted with a friend at EMC and he confirmed that the use of collections is encouraged.  :)

[Cheryle_Weather]

Sweet!

Thank you! and you, too, Joe!

[Cheryle_Weather]

"Additionally if you require DOD wipes or need to be 100% certain that the data has been destroyed upon deletion than you should not use collections.  Collected files that are "deleted" technically still exist in the collected object until the object itself is eligible for deletion."

One additional question regarding this statement.  If I'm understanding correctly, if we deleted a shortcut and the associated item, the associated item (now located in a clip) would not actually be deleted?  If this is the case, then, technically, the item would no longer be retrievable because the shortcut had been deleted?  I'm so confused :)

[Liam_Finn1]

ok here is how clips work

Inside a clip you have blobs (binary large objects) this represents your file or part of your file.

When you archive without collections each items is stored as a blob on centera

When you archive with collections many files are rolled into a c-clip (up to a max of 10MB) then the c-clip is stored on centera

When you do a delete without collections the item is deleted from Centera

if you do a delete with collections the c-clip and all it's contents remain on the centera until all items in the c-clip are marked for deletion. Then and only then is the c-clip and all data within the c-clip deleted

Example

File created in 2002 and a file created in 2009 are archived at the same time and both end up in the same c-clip

If you delete file from 2002, the reference for the item will be erased in EV but the actual data will remain on the centera.

Now you decide to delete the file from 2009. Since the file for 2009 and the file from 2002 are the only two files in the c-clip then the whole c-clip is deleted from the centera because all items contained within the c-clip have been marked for deletion

How this works will also depend on which version of centera you have

Basic

Compliance

Compliance Plus

Each version has different ways to enforce compliance retention

[Cheryle_Weather]

Ok... getting a little better grasp of this... sorry for being so thick!

"File created in 2002 and a file created in 2009 are archived at the same time and both end up in the same c-clip

If you delete file from 2002, the reference for the item will be erased in EV but the actual data will remain on the centera."

there is no way to retrieve the "deleted" item from 2002 with the reference being deleted from EV?

[Liam_Finn1]

Yes and No

Yes the item remains on Centera

But partially no on the retrieve

You will be unable to retrieve the data using EV BUT......

If you have the Clip-ID and the Saveset ID of the item then yes you can retrieve it using the centera API.

Chances are you wont have these two items but in theory the data is still available and retrievable if you have the above two items until the c-clip is deleted