Topic Last Modified: 2006-04-21

The Microsoft® Exchange Server Analyzer Tool uses the Exchange Server User Monitor (ExMon) tool to determine whether a user or a process (or a group of users or processes) is the cause of a disproportionately large percentage of the current server load.

There are several scenarios that ExMon evaluates:

The following sections provide details about and recommended actions for each scenario.

Client Restrictions

In this scenario, the ExMon RPC data indicates that client computer restrictions are causing a high load on this server. In the context of MAPI, a restriction refers to the MAPI operation called Restrict. This operation is used to select items that match particular criteria. The Restrict operation is part of the process to request that Exchange creates a view on a folder or set of folders (effectively, a database table with associated criteria). If the view on the folder or set of folders already has a matching restriction, Exchange uses the existing view to satisfy the user request. If a view does not have a matching restriction, Exchange creates a new view. Creating a view is more costly than using an existing view.

After many mailbox moves, client restrictions will be more costly for a short time. The move-mailbox operation causes a server to rebuild indexes on folders. Therefore, when clients view folders for first time on new server, all restrictions must be created anew and, as a result, performance may be poor. If you have not moved mailboxes recently, investigate that the following:

  • Another MAPI application is not installed on the client computers. Frequently, Microsoft Office Outlook® add-ins or other MAPI applications may cause too many restrictions to be created.

  • Mailboxes are sized and limits are set appropriately throughout your organization.

For more information about mailbox size and item numbers, see the Microsoft Exchange Team blog, "Recommended Mailbox Size Limits" (http://go.microsoft.com/fwlink/?linkid=55097).

High Bulk RPC Operations

Examples of high bulk RPC operations include moving, deleting, or setting properties (that is, the read flag) on hundreds or thousands of items in one operation. These types of operations may cause poor server performance for a short time. If you perform high bulk RPC operations frequently, and the server becomes bottlenecked at the disk or the processor, Outlook performance for the other users on the same server will be decreased.

To correct this error
  1. Verify that mailbox limits are set.

  2. Recommend that your users archive items to .pst files.

    Note:
    As a general best practice, folders that are used frequently should contain fewer than 5,000 items.

High RPC Synchronization Operations

High RPC synchronization operations indicate that many users are currently synchronizing items. As a result, server load is increased.

To correct this error
  • Make sure that you have scaled your Exchange servers according to how many client computers are running Outlook in Cached Exchange Mode. For more information, see Exchange Server 2003 RPC over HTTP Deployment Scenarios (http://go.microsoft.com/fwlink/?LinkId=47577).

  • Slowing down Cached Exchange Mode client requests may help prevent high RPC synchronization operations from becoming a performance problem. For more information, see Microsoft Knowledge Base article 818484, "Registry Values and Counters That You Can Use to Configure and Monitor Outlook Client Synchronization" (http://go.microsoft.com/fwlink/?LinkId=3052&kbid=818484).

Large Number of Client Online Operations

Sometimes, Exchange may experience an unusual spike in performance problems caused by high load that is distributed across users. This may be the result of poor planning as an organization grows. If you followed Microsoft best practices when scaling your organization, but you still experience poor performance caused by many client online operations, this may indicate that your users are running third-party MAPI applications or Outlook plug-ins.

User View Operations

ExMon RPC data indicates that a user is building new views, which causes high load on the server. In this scenario, a single user accounts for a significant percentage of the RPC CPU utilization.

To investigate this issue, consider the following scenarios:

  • The user may be performing operations on a folder that contains many items. To correct this problem, encourage users to reduce the number of items in any single folder. As a general best practice, folders should contain less than 2,000 items.

  • The user may be running a poorly written application that is causing excessive load. To correct this problem, remove the application.

For More Information