Video Screencast Help

Multiple Vault Stores

Created: 08 Jun 2006 • Updated: 21 May 2010 | 12 comments

We have 4 Exchange storage groups and was wondering if it makes sense to
create 4 EV vault stores to match the Exchange storage group. This will
allow us to logically separate the archives. Since single instance on EV
is on the vault level and Exchange is on the storage group level i can't see any disadvantage of doing this.
Are multiple vault stores recommended and what are the disadvantages of
this ?

How does multiple vault stores impact on EV journaling ?

Enterprise Vault 6 SP2
Windows 2003


Discussion Filed Under:

Comments 12 CommentsJump to latest comment

Jason Szeto's picture

In EV single instance is done on the Vault Store level. So having multiple vault stores you will lose single instancing between them.

Also it would help to know how you're using EV. A general rule of thumb is each EV component gets it's own vault store. Journal Store, Mailbox Store, PF Store, FSA store..etc.

Micah Wyenn 2's picture

I agree with jason on this one...but I'd take it even a step further and recommend a time based AND component based vault strategy. That way you've got an intuitive way to figure out when to dump vaults for new vaults (cause eventually they'll get big), as well as good intuitive searchability for users/DA users.


TonySterling's picture

That way you've got
> an intuitive way to figure out when to dump vaults
> for new vaults (cause eventually they'll get big), as
> well as good intuitive searchability for users/DA
> users.

Now that index volumes are being utilized if you set the AVSMaxLoc registry key the index volumes will roll over at a reasonable level and keep the indexes smaller. This will allow the CA\DA searches to be effecient and you don't have to dump the journal vault.

However, some companies will create new archives by the quarter, but they tend to get an extremely high volume of mail.

Micah Wyenn 2's picture

I wasn't really talking about the search times coming back slow. My experience with CA/DA is that the searches are normally pretty efficient given the size. I think I was more worried about the SQL db's homing the vault stores getting unweildly. As ya noted, a lot of our enterprise customers get a veritable shiznit load (that's the scientific phrase) of mail. Even at 400b per message, it adds up, especially when you're talking about journaling.

I've seen medium sized companies do 10gb/day journal...and that was a slow day (a weekend even). The db for that vault store was a couple of gig in a week (which made for an unhappy dba). *shrug*

Although now that I ponder it (or ramble upon it, as the case may be), does it make sense to seperate the journal vault for security, ease of searching from CA/DA, or whatnot? Oh, well, I usually go with a seperate machine for CA and journaling anyhow, and I usually home the vault store for journalling local the box anyhow (so I guess the best practice arguement is kinda mute for me anyhow). I wish I could say it was for speed, but really I just didn't usually want to put that much I/O over the network.

The AVSMaxLoc key is new fer me tho, school me on what the recommended values should be for a rollover setting?

TonySterling's picture

I normally set the key to 500,000,000. This will give you reasonable index volume folders.

Co Tran's picture

Since Exchange does single instance on the storage group level how does EV know it's the same message when it's archived using 1 vault store ?

eg. I send an email to 4 people who are on different storage groups, Exchange will have 4 copies of the email (1 on each SG). When EV archives that email does it keep 1 or 4 copies on the vault store ?

We are using mailbox archive and journaling only.

Micah Wyenn 2's picture

And we should be setting this in most to all installs? Or is this a special case sorta thing?


Micah Wyenn 2's picture

forums are broken again. :(Message was edited by:
Micah Wyenn

Micah Wyenn 2's picture

and here's another dup...but the next one should work.Message was edited by:
Micah Wyenn

Micah Wyenn 2's picture

The really really really simplistic overview answer is that as long as it's in the same vault store (no matter where the message came from/to/cc'd/etc), and it's "identical" then it will only store the message once the first time, and leave pointers to the rest that follow.

Now I say simplstic overview because SIS is actually kinda complex. It happens at a partition level, and it's hard to know when a message will actually be identical (as seen in the case of clustered exchange servers generating different msg envelope header information). SIS can also be configured to be disabled by accident, as it's option header is (at least in my opinon) somewhat misleading.

But in the real world, that isn't what you should really care about. After all, if we accept that attachments and messages are archived/indexed somewhat seperately, then the mail messages themselves don't really concern us all that much. What should make you smile is that the attachments will almost certainly get SIS'd, as the metadata about the attachments (binary or not) usually don't change. SIS loves binarys and will grab them most of the time. The win here is that your vault store physical storage usage will usually be 20% to 50% smaller due to SIS (and not to mention the dvs compression, but that's a topic for another post). 'course, don't make predictions based off that thumb in the sky stat, your mileage will vary according to the uniqueness of your site.

Besides, 23.872% of all statistics are made up on the spot. Oh, and 7 outa 5 people have trouble with fractions. I guess they should teach more math here in the good 'ol USA.


TonySterling's picture

I believe the SIS for EV is determined by a checksum on the message header.

Micah Wyenn 2's picture

Where's the AVSMaxLoc key live? It's not described in the registry.chm nor is it in the demo images, or docs. :) Help a brutha out. :)