Scaling EV
Is there a point when it is recommended to split off the Mailbox Archive and Journal Archive into separate EV directories (Approx 24K users)? My Client is looking to use DA and CA against the Journal. Is there a published doc that references best practices for such decisions?
No, there isn't going to be a doc designed for public consumption. It really depends on the volume of mail to be archived\journaled.
Did you have your EV system designed by a certified Symantec partner? I am a bit surprised you don't have a dedicated journal server already but again, it comes down to the amount of mail.
Do you have two vault stores, one for mailbox and one for journal?
yes that's correct and i'm sure tony was going down the same path before i interjected. in certain cases with extremely large or geographically diverse environments there could be circumstances which would dictate seperate EV directories but i dont think you're there. 25k is large but by no means the largest in terms of EV's scale.
assuming all your indexes are 64bit already... if you run into performance issues with DA you could explore the option of seperating DA's SQL databases onto a dedicated server, increasing the indexing resources on your journal servers, etc. you might even look into Clearwell as an option depending on your client's eDiscovery requirements.
I would strongly recommend that you archive all journal mailbox data into one retention category and all user mailbox data into another retention category. This is a pretty simple change to make by just reconfiguring the properties of the Journal target, provisioning groups, etc. You will still be able to single-instance between vault stores to whatever degree you are currently doing that. It's not an uncommon request for the business to want to change their retention strategy, and handle different data differently; for example, you may retain everything "forever" currently, but the company rules change and you want to expire journals after 7 years and user data after 3 years. If everything is in one big bucket (retention category) this is a nightmare but if each of these two distinct data sets is in its own category, it's about 4 minutes of work to make the change.
You can have two retention categories that are confgured to retain data for the identical amount of time. This just allows you easy flexibility down the road. If you never use it, you're only out a few minutes of admin time that it took to set up.
As for how big a site can be before it needs to be split, that is kind of a "how long is a peice of string" question but from a purely mechanical standpoint, I know of one customer with about 55 EV servers and 140k users, mix of mailbox and journaling, all in a single site. The only challenge their architecture has created is for their once-or-twice-a-year upgrades which become an "all hands on deck" kind of thing; all EV servers in the site must be upgraded in the same window (can't upgrade 4 today and 5 tomorrow unless you don't mind none of them working till they are all done) so they need a lot of hands to get them all done. They prefer this because it makes their daily admin work a lot more streamlined since it is all in one site.
