Hi, I ran the Exchange 2010 Mailbox Server Role Requirements Calculator and there is something I don't understand, disk-wise, about the results for a particular scenario. It seems to me that too many disks are needed compared to what I expected. I don't doubt the results, I just hope someone will be able to explain to me the reason! The scenario is basically a DAG, with 2 copies and 1 lagged copy, for 2000 mailboxes with a quota of 200 MB. I also chose to implement a Restore LUN. Here are some input that I gave it which I think is what you need to understand my scenario. Let me know if you need something else: INPUT Number of Mailbox Servers Hosting Active Mailboxes / DAG: 2 Total Number of HA Database Copy Instances (Includes Active Copy) within DAG: 2 Total Number of Lagged Database Copy Instances within DAG: 1 Dedicated Maintenance / Restore LUN?: YES Total Number of Tier-1 User Mailboxes / Environment: 2000 Mailbox Size Limit (MB): 200 Disks for Database: 2000 GB, 7.2K SAS 3.5" Disks for Log: 2000 GB, 7.2K SAS 3.5" Disks for Restore LUN: 2000 GB, 7.2K SAS 3.5" RESULTS The Role requirements tab gives me this: Disk Space Requirements /Server Database Space Required 748 GB Log Space Required 664 GB Database LUN Space Required 1028 GB Log LUN Space Required 830 GB Restore LUN Space Required 1764 GB In the Storage Design tab, I am proposed RAID-1/0 for the Databases. I entered 0% for the RAID-1/0 Rebuild Overhead, in order for the rebuild overhead not to have any influence on the number of disks I need. So for the 1028 GB LUN on a server, I thought I would need 2 X 2000 GB disks because it's RAID-1/0, and I only have 1028 GB of data. But the calculator says that I need 6 disks per server, which is 4 more than I thought. That is what I don't understand... Somebody could please explain this to me ? I'm sure there is another factor, but I can't find it. Thank you.

You're only looking at it from a pure space perspective, and wven then you're not accounting for the dedicated restore LUN. Try looking at it from a performance perspective. John

Hi Here is example for exchange 2010 Mailbox server design. You can calculate capacity requirements by yourself. http://technet.microsoft.com/en-us/library/ee832789.aspx It is the blog of calculator. http://blogs.technet.com/b/exchange/archive/2009/11/09/3408737.aspx Maybe they are helpful to you. 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.

You're only looking at it from a pure space perspective, and wven then you're not accounting for the dedicated restore LUN. Try looking at it from a performance perspective. John Hi John, You're right, I wasn't looking at it from a performance perspective. I didn't think as much as 4 extra disks would be needed just to give me the needed performance. With Microsoft saying that I/O between Exchange 2003 and Exchange 2007 was reduced by 70%, and another 70% from Exchange 2007 to Exchange 2010, I expected way smaller requirements from a performance perspective. I am thinking of going below those requirements, as these seem to be worst-case scenarios. With JetStress and LoadGen, I could test to see if going below those requirements makes sense. What do you guys think of that? As for the dedicated restore LUN, I do account for that restore LUN. The 6 disks that I was talking about are 6 disks per server in RAID-1/0 for the Database only. The calculator also suggests me 2 disks per server in RAID-1/0 for the Logs and 3 disks in RAID-5 for the Restore LUN. Thank you.

Remember that the number of databases plays here as well. You want to reduce the number of databases to the minimum possible while staying within the Microsoft guidelines for maximum database size. This is because each database adds a number of non-transactional IOPS for things like Online Maintenance and Indexing. It's around 30 IOPS per database. With one 2TB database, the non-transactional IO is 30 IOPS. With 10 200gb databases, the non-transactional IO becomes 300 IOPS. It makes a big difference when using SATA spindles. John

Thank you John for pointing me to the performance factor, and thank you Terence for the pointer to the calculation example.