Forum Discussion

KeirL's avatar
KeirL
Level 6
12 years ago
Solved

EV Index sizes and locations for EV10

I'm anticipating EV creating some large indexes due to a migration project of a significant number of PST files. Indexing is likely to be about 30TB. There will be several EV servers managing this migration activity

How should I plan my index volumes - I could create several 20TB index locations (one for each EV server) to cope with the migration and on-going steadystate, but is this too large.... Perhaps I would be better to create a larger number of smaller volumes (say 5TB??? or even 2TB??) and then actively manage the closing of these index locations as they become full and opening new locations.

Which would be more efficient for subsequent searches and e-discovery

many thanks

Enterprise Vault
Information Governance
Windows Server (2003-2008)
  • GertjanA's avatar
    GertjanA
    12 years ago

    Remember that in EV10 a closed indexlocation is really closed. That location could be taken out of regular backup once closed, and back up once a month.

    What is your backup tooling capable of doing? I think that this should be the limiting factor. If you can (for instance) backup an index disk of 5TB during your backup window, I would go for that size.

    Then use VM'P's for the index-locations.

     

     

6 Replies

  • Rob_Wilcox1's avatar
    Rob_Wilcox1
    Level 6
    12 years ago

    Given that full indexing is about 10-15% of the data that is being indexed... are you sure on the numbers? Are you really going to have about 300 Tb of archived data?

  • KeirL's avatar
    KeirL
    Level 6
    12 years ago

    Hi Rob

    Yeah - that does include several years growth but I'm going by the results from EMA and the sizing spreadsheet - it's a large environment.

    Assuming I'm going to need to provision some pretty large index locations what do you think would be the best strategy - I really just trying to guage how indexes need to be factored into the volume size calculations. They still need to be backed up so should they follow a similar strategy to VS Partitions whereby they can be closed when they get to say 1TB and backed up less frequently?

  • GertjanA's avatar
    GertjanA
    Moderator
    12 years ago

    Remember that in EV10 a closed indexlocation is really closed. That location could be taken out of regular backup once closed, and back up once a month.

    What is your backup tooling capable of doing? I think that this should be the limiting factor. If you can (for instance) backup an index disk of 5TB during your backup window, I would go for that size.

    Then use VM'P's for the index-locations.

     

     

  • KeirL's avatar
    KeirL
    Level 6
    12 years ago
    Thanks both
     
    We're actually using NetApp so backups will be via snapmirror to the DR site - so it sounds like 5TB is a good size before closing indexes and starting over?
     
    Gertjan - I'm still interested on what you do with your indexes if you're using building blocks..... can you take a look at my thread below and comment accordingly :o)
     
  • GertjanA's avatar
    GertjanA
    Moderator
    12 years ago

    Hello Keir,

    Our setup is a bit 'off'. I've taken over this environment, so this is NOT my design...

    We have 3 live servers (ev001, ev002, and ev003), and 1 DR-server (EV301)

    Indexes are located as follows:

    EV001 - F:\Enterprise Vault\Indexes\EV001\Index-VMP01 (and an Index-VMP02, and an Index-VMP03)

    EV002 - F:\Enterprise Vault\Indexes\EV002\Index-VMP01 (and VMP02, VMP03)

    EV003 - F\Enterprise Vault\Indexes\EV003\Index-VMP01 (and VMP02, VMP03)

    These locations are all replicated to the EV301, which has:

    EV301 - F:\Enterprise Vault\Indexes\EV001\Index-VMP01 (and VMP02, VMP03)

    EV301 - F:\Enterprise Vault\Indexes\EV002\Index-VMP01 (and VMP02, VMP03)

    EV301 -  - F:\Enterprise Vault\Indexes\EV003\Index-VMP01 (and VMP02, VMP03)

    Because there is a differentiation (if that is the proper word) on foldernames )indexes\EV00x etc), when a DR is being done, the EV301 server works on the 3 main indexlocations. If I failover 1 server only, the DR-server works on 1 folder. Obviously, this requires some additional config in replicating the data back. If we do a DR test, we keep EV in backup mode to prevent changes, and to prevent the need to replicate data back. If we have a real DR, replication obviously is being done the other way. (if available...) The same applies for the data, but as you have netapp, that is handled by netapp config (revert replication?).

    I'll see if I can compose a document, to show the scripts.