My favorite points considering planning / deploying for Exchange 2010 storage design.
- Storage capacity of the HDD are getting increased by ages
- The RPM/IOPS are somewhat still constant or an average to result performance
- You need to look at the data protection management strategy
- You need to look at the risk management strategy using discovery retention policy
- You need to look at user experience strategy as client performance – it should be able to match the growth of storage
- You need to look at online moves & upgrade – You cannot keep the mailbox for 10 years as it is & would want to maintenance
- You need to look at the engines…like fast searches give the results very fast using single or multiple mailboxes
- Larger database cache size & compression
- 8KB (Exchange 2007) to 32Kb (Exchange 2010) page file size
- Re-Architect the exchange storage schema as the disk size increase
- Multiple copies of exchange databases using DAG
- IOPS reduces 70 percent from Exchange 2007 & 90 percent from Exchange 2010.
Larger mailboxes one of the most favorites to have maximum size …when related the topic with the disk subsystem since exchange 2007 had high cost storage IOPS utilization the size was limit but due to the drastic IOPS reduction and the less expensive disk the performance was pulled out and the mailbox size was increased…!
Exchange 2010 combines multiple IO operation into single IO operation resulting in to single IOPS instances which decreases the frequency of disk IO operation, reduces the IOPS & Improving performance
To increase performance and reduce IOPS in exchange 2010 from exchange 2003 & 2007 the data are read/write from the database to and fro having the larger page file size. I.e. 32k
The Exchange database has tree structure binary hierarchy and every branch/page is of 32KB read/write which increases the performance and reduces the IOPS.
In Exchange 2007 data are written in databases randomly therefore it requires more IOPS instances to read/write data.
ESE in Exchange 2010 writes data sequentially to the database resulting related data stored very nearby under the binary tree which results in few IO instances.
B-Tree de-fragmentation process in Exchange 2010 has new three operations as below: which reduces IOPS utilization. : Page move, Partial left merge & Full left merge.
Exchange 2007 Database defragmentation used to affect the performance. In Exchange 2010 contains a new event which takes care of the background defragmentation which runs across the B-Tree and identifies the areas which needs to be defragment. It complies in the index and defrags in the back ground so the load on exchange server is less which increases the performances and eliminates the downtime maintenance.