Enterprise Vault building block

[ecrgvp]

Hello team,

I have seen several customers running indexing service on a separate EV server dedicated for this purpose. Can someone advice on the merits and demerits of running index services on a separate server?

Also, is there a document which explains on what services can be run on a separate dedicated EV servers and what are the prerequisites and considerations.

Thanks

V

[JesusWept3 · Accepted solution]

You can push out all sorts of things to different servers if you wanted to
so if you wanted to go nuts you could have

EVServer1 - Task Controller that connects to your mailboxes
EVServer2 - Storage Server that controls your Vault Stores and Vault Store Partitions
EVServer3 - Indexing Server that controls your indexes

I'd never recommend that above, but you *could*

The biggest "con" against having the indexing server seperate is that as it builds the indexes you will be making more work for your network, so if it calls a bunch of items to index, it will go off to the Storage Server, recall the DVS Parts from the storage locations (perhaps more network traffic its network attached) and then passes those items across the network to the indexes

So this will have performance impact on how fast Indexing can get a hold of the items and how much network traffic will be taken up by the entire process

The thing is though when Enterprise Vault is converting items to Text or HTML, the EVConverterSandBox.exe can take a lot of CPU time, and the more processer time thats being taken up by other processes means that other processes such as the IndexService will take a back seat, meaning indexing *could* be slower

The issue though really comes down in to this being another moving part that can fail, the more local everything is, the less chance of failure

Though in the future this could all change, but for now, There aren't many scenarios (if any) that i could really recommend the indexing server being seperate from the Storage and Task Controllers

 

[WiTSend]

Offloading the Index service might give you some additonal internal resources for the Indexing process, but it is not generally considered best practice.

1)  If the index service were to stop all of the archiving servers would be affected

2)  This creates a lot of MSMQ traffic running on the WAN between the various EV servers.

By using the building block approach, each EV server is essentially self-contained, doing it's own archiving, indexing and storage.  Failure of any one server (or service) only affects the users associated with that particular server.  Failover (using USL) is simple and effective.

Just my 2 cents...

[JesusWept3]

Indexing won't cause an increase of MSMQ traffic at all.