Hi, we are moving to Exchange 2010 SP1 from 2007 SP2. I am planning the SAN allocation and have an idea that I would like to check for compatibility. We have 2 production datacentres, let's call them East and West. In each datacentre we have a SAN that has Tier1 SAS disks and Tier 3 SATA disks, and we have the ability to change the tiering of LUNs whilst the data is online. I am planning one DAG that has 2 mailbox servers in each datacentre. I am planning that each mailbox server will have it's own C drive, but the 2 mailbox servers in the East datacentre will have their active mailbox databases in a LUN with drive letter E: that is on the high speed SAS disks. Then all the passive mailbox replicas from the West datacentre will be on drive W: which will be a LUN from the SATA disks. Then, in the West datacentre, the 2 mailbox servers will each have their own W drive too, but these will have the active mailboxes on the high speed disks and the E drive will be for the passive copies on SATA disks. Will this work or be advisable?

Hi, Thank you for your question. I am trying to involve someone familiar with this topic to further look at this issue. There might be some time delay. Appreciate your patience. Gen Lin TechNet Subscriber Support in forum If you have any feedback on our support, please contact tngfb@microsoft.com Please remember to click Mark as Answer on the post that helps you, and to click Unmark as Answer if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Thanks Gen Lin-MSFT

Another thing in your setup is that you'd want the active databases to survive a WAN loss between the two datacenters, and as such you'll need to create 2 DAGs with the File Share Witness placed in the same datacenter as your active databases to maintain quorum if the network conectivity fails. Take a look at: Two DAGs for the active/active user distribution model here: Understanding High Availability FactorsJesper Bernle | Blog: http://xchangeserver.wordpress.com

I look forward to an answer, thanks.

Thanks, I looked at two DAGs, but here is my architecture; East has 250 users on-site. West has 400 users on-site. There are 1850 users who are all remote to the datacentres, our head office ("North") has no IT infrastructure other than a single local fileserver/DC, and has 300 staff. We also have another site ("South") that has 500 users, but again no local Exchange server. Currently we have 2 Exchange 2007 servers, with 10 mail store databases with around 125 users on each, to keep each DB to below 100GB. We have distributed the "remote" users across the datacentres, where most of "North" have their mailbox in "East" and most of "South" have their mailbox in "West". We have a DR RTO of 1 week, but aim for less than 3 days outage. We want to exceed the business's expectations and bring this down to minutes or hours. East has 6 links that can go to West. So, there are potentially 5 backup routes between the datacentres (effectively routing through "south" for example), so the risk of a inter-Datacentre WAN loss is very low, and the number of affected users (on site people) will be comparitively low. We also have a dedicated link between East and West for replication, and this can be used too. Between datacentres, latency is 13ms. Other sites connect to the Datacentre with around 25ms latency at worst. So, to support the remote users who outnumber the on-site users, we have planned for a single DAG with 4 mailbox servers, and all mailbox servers having a copy of every database. To achieve this, I want to create two disks, where the local databases are on faster disks, and the passive copies are on cheap disks - but to keep the drive letters consistent this means that the drive performance will differ in each site. In the "East" datacentre, the E: drive will contain all mailbox databases for people who are homed on those two servers. The W: drive will have passive copies from the West datacentre and this will be a couple of SATA disks. In the same way, in the "West" datacentre, the W: drive will be on a SAS RAID array with the active databases for the users in that location and geographically close locations, and the E: drive will be on some SATA disks that will contain the replicas from the mailbox servers that are in the "East" datacentre. Please let me know if I have made any errors in my analysis that points to a different layout such as 2 DAGs or different disk architecture.

Well, it sounds like you can manage with a single 4 node DAG. Since bringing in another DAG into the solution would mean alot of extra Exchange servers and costs, my recommendation would be to try it out and if you find it doesn't meet your requirements it's easily scaled out. Some of the choices one has to make in designing a DAG isn't always so simple to make since it's all about balancing cost versus features to reach the desired ROI objective.Jesper Bernle | Blog: http://xchangeserver.wordpress.com

I am really hoping that a 4 node DAG will meet all of our performance needs, because only 2 years ago we had a single Exchange 2003 server (with 2 CPUs and 4GB RAM) handling 90% of the same load as we have now. With Exchange 2007, we scaled this to 3 servers, and with Exchange 2010 we are scaling this to 8 servers (4 of them CAS/HT). In my planning, I cannot justify 4 Mailbox servers with 12GB in each of them - so I am planning to make them smaller at 8GB each - and then scale them larger if needed. Has anyone tried 8GB for 750 users (36 sent/received a day average 47Kb) and experienced any problems?

Has anyone tried 8GB for 750 users (36 sent/received a day average 47Kb) and experienced any problems? For 50 MSGs @ 75 KB per user the database cache per user is roughly 3 MB. And per 750 users that would be 750 x 3 = 2250 MB. And given you should give the Mailbox role 4 GB + the User load that would give you 6,25 GB. So 8 GB should give you ok performance. Understanding Memory Configurations and Exchange PerformanceJesper Bernle | Blog: http://xchangeserver.wordpress.com