XGEN: MTA Queue to Information Store Processing Slowly (234702)


The information in this article applies to:

  • Microsoft Exchange Server 5.5
This article was previously published under Q234702

SYMPTOMS

If you view the message transfer agent (MTA) queues by means of the Exchange Server Administrator program or Performance Monitor, the information store (either public or private, or both) may exhibit a backlog in the message queue. This behavior is often evident on servers with two to four processors and 800 megabytes (MB) or more of RAM. Manual tuning of the registry allows full use of hardware resources and more efficient clearing of the message backlog under high-traffic conditions.

CAUSE

Because computer hardware has become less expensive, many companies use more RAM and processors on Exchange Server computers to increase performance. However, Performance Optimizer (Perfwiz.exe) does not automatically tune some registry parameters to fully use all of these hardware resources.

You can confirm this behavior in Performance Monitor through the counters specified below:

Object: MSExchangeMTA Connections
Counter: Queue Length
Instance: Microsoft Private MDB
Description: Queue Length is the number of outstanding messages queued for transfer to the entity.

Object: Private Information Store
Counter: Receive Queue Size
Description: Receive Queue Size is the number of messages in the private information store's receive queue.

If these two values are more or less equivalent, the information store is not effectively processing the inbound transactions from the MTA, and the bottleneck is a result of the information store's inability to handle these transactions in a timely manner. Also, the MTA may not be handing off mail quickly enough for the information store. In addition, MTA and information store CPU use (as monitored by Performance Monitor) may be low on the affected server.

WORKAROUND

To work around this problem, increase resource (thread) settings from the defaults:
  1. From a command prompt go to the Exchsrvr\Bin directory, and run perfwiz -v to start Performance Optimizer in verbose mode. Verbose mode stops all services and offers options not available in normal mode. You do not need to analyze hard disks or move files; this procedure presents the opportunity to change some resource (thread) settings from the defaults. In the Microsoft Exchange Performance Optimizer property page, in the Users on this server and Users in organization areas, be sure you allocate enough resources to handle your work load. If your number of users falls in the high end of a range, select the next (higher) range. In the Memory Usage area, don't limit Exchange Server memory usage.

    To view the current settings without stopping services, at a command prompt, run perfwiz -r. You can also use the Perfopt.log file in the Winnt\System32 folder to view any changes that Performance Optimizer makes to the system.
  2. In Performance Optimizer, click Next until the # of background threads property page is displayed. Note this value, but don't change it. Click Next.
  3. In the Public and Private Information Store property page, increase each of the four send and delivery thread values to 8.
  4. Click Back to move back to the # of background threads property page, and increase that value accordingly. For example, if you increased each of the four send and delivery thread values from a default setting of 2 to 8 (by 6 threads each) in step 3, then in the # of background threads property page increase the background threads value by a total of 24.

    You can make these changes manually to the registry, but these values are not visible by default. Running perfwiz -v automatically adds these values for you and eliminates any worry about the spelling or case sensitivity of a value.
  5. Click Next until the # of submit/deliver threads property page is displayed. The MTA uses these threads to pass mail off to the information store. Increase this value to 8 also; its maximum default setting is 3.
  6. Click Next, click Finish, and then restart services. Monitor the MTA queue to the information store to verify whether the changes have any effect on the queue.
Increasing these threads' value to 8 allows for a larger "pipe" between the MTA and the information store, and more efficient clearing of the message backlog under high-traffic conditions.

For additional information about this problem, click the article numbers below to view the articles in the Microsoft Knowledge Base:

230691 XADM: Space Tree Fragmentation in Exchange Database Engine Can Degrade Server Performance

195006 XADM: Resource Failure Error on Computers with More Than 800 MB of RAM

Modification Type:MinorLast Reviewed:4/28/2005
Keywords:kbprb KB234702
©2004 Microsoft Corporation. All rights reserved.