Chapter 2: Capacity and Availability Management 13
resulted in specialized hardware for specialized functions, for example, in the case of front-
end servers.
To ensure that you manage capacity appropriately for your Exchange 2000 server, you
need a great deal of information about current and projected usage of your server running
Exchange. Much of this information will come from monitoring. You will need informa-
tion about patterns of usage and peak load characteristics. This information will need to be
collected on a server-by-server basis, because a problem with a single server in an Exchange
2000 environment can result in a loss of performance for thousands of users. The perfor-
mance of your network is also critical in ensuring delivery times and timely updating of
Exchange directory information.
The main areas you should monitor to ensure that your servers running Exchange exceed
your SLAs’ capacity requirements include the following:
◆CPU utilization
◆Memory utilization
◆Hard-disk space used
◆Paging levels
◆Network utilization
◆Delivery time within and between routing groups
◆Delivery time to and from foreign e-mail systems within your organization
◆Delivery time to and from the Internet (although this depends greatly on minute-by-
minute performance of your connection to the Internet and the availability of band-
width to other messaging environments)
◆Time for directory updates to complete
You will find more information on monitoring in Chapter 4.
It is fairly common to choose the size the disks of a server running Exchange based on how
many mailboxes you plan to have on the server multiplied by the maximum allowable size
of each mailbox. Using this approach, however, will generally not help you to meet your SLAs.
You should strongly consider approaching this problem from a different perspective. When
determining the capacity of your server running Exchange, consider basing it on the time it
takes to recover a server from your backup media. Recovery time is generally very important
in organizations because downtime can be extremely costly. If you are using a single store
on your server running Exchange, use the following procedure to help you size it.
1. Divide the recovery time of a database (defined in your SLA) by half. Around half the
recovery time will generally be spent on data recovery, the rest on running diagnostic
tools on the recovered files, database startup (which includes replaying all later message
logs) and making configuration changes. Of course, this is only a general figure—the
longer you leave for recovery time, the smaller the proportion of that time is required
for configuration changes.
2. Determine in a test environment how much data can be restored in this time.