Storage improvements in Exchange Server 2010 allow you to consider more inexpensive and slower types of disk storage, which you might not have been able to consider for previous versions of Exchange Server. However, ultimately you need to test the storage configuration that you select to ensure it meets your needs.
SATA disks typically have a slower random access disk performance than SCSI or serial attached SCSI (SAS) drives. However, since Exchange Server 2010 has optimized performance to increase the use of sequential disk access, you can use SATA drives in a wider range of scenarios. SATA disks are significantly less expensive than SCSI or SAS disks of the same storage capacity.
Direct-Attached Storage (DAS) is significantly cheaper than using a Storage Area Network (SAN), but a SAN typically performs better than DAS, and is more reliable. However, when you use a database availability group for high availability, the need for reliable storage is reduced, and you can consider DAS as an option.
An Internet SCSI (iSCSI) SAN typically has lower performance than a Fibre Channel SAN, but it is also significantly less expensive. The lower I/O requirements in Exchange Server 2010 make iSCSI an option over Fibre Channel in a wider range of scenarios when a SAN is desired.
A variety of RAID types are available to increase performance and redundancy of the disk system. RAID 10 is the best performing RAID option, because it has the speed of a striped set and the redundancy of mirroring. However, it is fairly expensive, because 50 percent of disk space is wasted on redundant data.
JBOD is a recommended configuration if you have three or more replicas of a database in a database availability group. JBOD has no redundancy or increase in performance over single disks, but are expandable because additional disks can be added. If disks fail, another replica in the DAG can begin servicing requests. This reduces the need for redundancy in the storage system.
To ensure that a Mailbox server meets your performance requirements it must be tested. The following lists considerations for creating a test plan.
Gather usage information
To design a test plan for Mailbox server performance, you need to accurately understand how the server will be used. This includes factors such as the number of mailboxes, the number of messages users will send, and the type of clients that will be accessing mailboxes. If you do not accurately understand the load that will be placed on the server, it is impossible to ensure that server performance meets your needs.
Replicate the production environment
When you create your test environment, you should ensure that it replicates the conditions in your production environment as closely as possible. This means that you should be using identical hardware, software, and drivers on the test system and production system.
Use performance testing tools
To test server performance, it is impossible to completely replicate the users in a production environment. However, Microsoft provides two tools that you can use to generate simulated loads on the server:
Exchange Load Generator (Loadgen). You can use this tool to create a simulated load of MAPI, Outlook Web App, Exchange ActiveSync®, Internet Message Access Protocol (IMAP), POP3, and Simple Mail Transfer Protocol (SMTP) clients on Exchange servers. You can configure this tool based on the usage data you have gathered to determine whether the performance is acceptable.
Jetstress. You can use this tool to verify disk performance by simulating the Exchange Server database and the log file loads that a specific number of users produce. It is also capable of simulating the load generated by database replication in a DAG.
In many cases, you can use virtualization to test the functionality of proposed configurations, including Exchange Server 2010. However, virtualization is not appropriate for testing server performance unless the production Exchange server is also virtualized.