Planning an Exchange Server 2003 Journaling Deployment


Topic Last Modified: 2005-04-27

Widespread journaling will have an impact on the performance of Exchange. Therefore, it is likely that you will have to deploy more hardware to maintain the current level of messaging service in your organization. Because every organization is different and compliance solution frameworks vary greatly, this section provides only high-level recommendations about performance analysis for server and topology sizing. It is recommended that final decisions be made from information gathered in a lab environment that is as similar as possible to your production environment.

Most of the CPU and disk input/output (I/O) consumed by journaling is the result of an increased burden on the store process on the journalized mailbox databases and on the servers hosting the journal recipient mailboxes.

Therefore, from a performance and server load perspective, there are two classes of server you have to size: the user mailbox servers that you want to journalize and the mailbox servers that host the journal mailboxes.

The following sections about planning discuss how to size these two classes of servers. First however, a brief discussion of how the surrounding infrastructure will be affected is necessary.

The main impact on infrastructure is the increase in network traffic because of enabling journaling. The impact is directly proportional to the number of users who are being journalized and how much mail is sent between mailbox databases in the organization. For example, if only a small part of users are journalized to satisfy the requirements of the Sarbanes-Oxley regulation, network impact will be much less than if an entire division is journalized to satisfy the requirements of the Sec 17a regulation. As for the impact on mail sent between mailbox databases, mail sent from a user to another user on the same journalized mailbox database generates only a single message to a journal recipient mailbox. However, if a journalized user sends mail to a journalized user on a different mailbox database with a different journal recipient, two messages are sent to two different journal recipient mailboxes. Therefore, mailing habits and how you group users on the mailbox databases in your organization affects the overall network load after journaling is enabled.