Exchange 2013 has many new features and changes in comparison to Exchange 2010. Here, I’m going to discuss one such change that involves both active and passive database copies. In Exchange 2010, IOPS for a passive database copy is the same as that for an active database copy. However, in Exchange 2013, IOPS for a passive copy is about 50% of that of an active copy.
In a balanced Exchange storage design, active and passive database copies are well distributed among the mailbox servers within a DAG (Database Availability Group). That means each mailbox server will have some number of active databases and the same or a different number of passive databases.
This creates a scenario where a mailbox server will have two different user profiles, or two different workloads, if you like, one for active databases and the other for passive databases.
How would you simulate such a scenario? No problem. You can run two instances of Jetstress 2013, simultaneously, on a single mailbox server.
Figure 1 below shows an example of two Jetstress user profiles with 3,000 mailboxes, 3 active databases and 3 passive databases. Note that for the passive database copy, the target IOPS is 50% (or half) of that of the active database copy.
Figure 1. Two Jetstress user profiles, one for active and one for passive databases.
Figure 2 shows the two instances of Jetstress running on a server simultaneously. The instance in background is for the active database copy, while the one in foreground is for the passive database copy.
Figure 2. Two instances of Jetstress running simultaneously, simulating both active and passive databases.
Figure 3 shows the header portion of the Jetstress 2013 reports, for both instances of Jetstress. For the active database copy (on the left hand side), the targeted and achieved IOPS are 180 and 313. And for the passive database copy (on the right hand side), the targeted and achieved IOPS are 90 and 146.
Figure 3. Two Jetstress reports (header portion).
Figure 4 shows the configuration portion and a fragment of the performance result portion of the Jetstress 2013 reports for both instances of Jetstress. Again, the report on the left hand side is for the active copy, while the one on the right for the passive copy.
Figure 4. Two Jetstress reports (configuration and results portion).
As you can see, the transactional read and write latencies are comparable for active and passive database copies. This is by design because both active and passive database copies are sharing the same pool of hard disk drives.
Thanks for reading.