Adjusting Software Settings

Using Performance Optimizer is an essential step in maximizing your server's performance. This utility selects the software settings that are appropriate for your configuration and places Microsoft Exchange Server files in the best locations. It configures your server on the basis of the following factors:

The number of users that the server supports and the number of users in your organization.
The types of activities the server is to perform.
The server's hardware resources, including the amount of RAM and the number and type of disks and processors. It is recommended that the amount of RAM remain at zero, which means that Microsoft Exchange Server will use all available memory. Note that this ultimately depends on the organization, as an organization might run another process on the same server.

For example, Performance Optimizer uses the number of clients that the server is to support to determine the number of MAPI/RPC threads that are to be allocated for the information store.

Run Performance Optimizer after you run Setup and when you change your server's hardware configuration. In particular, you should run Performance Optimizer after you do any of the following:

Make changes to your server's disk.
Add a caching controller.
Change the amount of RAM.
Add a processor.
Change the number of users that the server is to support.
Add or remove a component on the server.

You can customize Performance Optimizer settings by typing perfwiz -v at the command prompt in the Exchsrvr\Bin directory. When you run perfwiz -v, Performance Optimizer displays six additional screens that provide the options shown in the following table.

Option Description

# of information store buffers The maximum number of 4KB buffers allocated to the information store database.

# of directory buffers The maximum number of 4KB buffers allocated to the directory store database.

Option Description

Minimum # of information store threads The minimum number of threads that the information store will use to service Messaging Application Programming Interface (MAPI) clients, such as Microsoft Outlook.

Maximum # of information store threads The maximum number of threads that the information store will use to service MAPI clients, such as Microsoft Outlook.

# of directory threads The maximum number of threads used by the directory.

Maximum # of concurrent read threads The maximum number of directory threads available to service replication requests.

# of background threads The number of threads available for background tasks and to the Gateway In/Out, Send, and Delivery thread pool.

# of heaps The number of areas of memory used for dynamic memory allocation.

# of private information store send threads The number of threads that process messages submitted by MAPI clients, such as Microsoft Outlook.

# of private information store delivery threads The number of threads that the private information store can use to deliver messages to mailboxes.

# of public information store send threads The number of threads that process public folder replication messages and messages generated by public folder rules.

# of public information store delivery threads The number of threads that the public information store can use to deliver messages to public folders.

# of information store gateway in threads The number of information store threads that deliver mail from the message transfer agent (MTA) to the information store for routing elsewhere. For optimal performance, the number of threads should be increased only if the server has multiple processors.

# of information store gateway out threads The number of information store threads delivering mail from the information store to local mailboxes, or to the MTA. For optimal performance, the number of threads should be increased only if the server has multiple processors.

Buffer Threshold Low Percent The percentage of available buffers remaining before buffers are flushed to disk. Lower values limit the number of writes to disk; however, low values can degrade performance.

Buffer Threshold High Percent The percentage of available buffers that must be reached before flushing of buffers to disk stops. For optimal performance, set the percentage value equal to, or slightly greater than, the Buffer Threshold Low Percent.

Maximum # of pool threads The maximum number of threads servicing Internet connections such as Internet Message Access Protocol, Version 4rev1 (IMAP4), Post Office Protocol version 3 (POP3), and Network News Transfer Protocol ( NNTP) into the information store. This value is per processor.

# of information store users The number of users for which this server routes mail.

# of concurrent connections to LAN-MTAs The number of concurrent network associations to MTAs.

# of concurrent connections to RAS LAN-MTAs The minimum number of concurrent network associations to Remote Access Service (RAS) MTAs.

# of LAN-MTAs The minimum number of network MTAs supported by the MTA.

# of X.400 gateways The maximum number of remote MTAs connecting through X.400 using X.25, Transport Control Protocol/Internet Protocol (TCP/IP), or Transport Class 4 (TP4).

ds_read cache latency (secs) The number of seconds before the directory service read cache expires. Performance improves when items are loaded in the cache by limiting the number of directory service reads.

# of dispatcher threads The total number of threads used to route messages.

# of transfer threads The total number of threads used to transfer messages.

# of kernel threads The total number of threads allocated to process the Open Systems Interconnection (OSI) protocol stack.

# of submit/deliver threads The total number of MTA submit and deliver threads. For example, if this value is set to 3, the total number of threads is 6. Submit threads receive messages from the information store and deliver threads deliver messages to the information store.

# of RAS LAN-MTAs The maximum supported number of concurrent network associations to other RAS MTAs.

# of database data buffers per object The number of 4 KB buffers configured per cached MTA database file. The MTA saves a copy of each message until the message has been accepted by the information store, or another MTA.

# of RTS threads The total number of threads available to the Reliable Transfer Service (RTS) level of the OSI protocol stack.

# of concurrent MDB/delivery queue clients The maximum number of information store and XAPI MA delivery queue clients supported. Each client can have more than one session. For optimal performance, the value should be at least 2, to support both the public and private databases.

# of concurrent XAPI sessions The maximum number of sessions to information store and XAPI MA delivery queue clients, XAPI MA retrieval queue clients, and XAPI MT gateway clients.

Max # of RPC calls outstanding The maximum number of concurrent remote procedure call (RPC) threads. This limits the maximum number of RPCs that will be processed at one time.

Min # of RPC threads The minimum number of concurrent RPC threads. This sets the minimum number of RPCs that will be processed at one time.

# of MT gateway clients The maximum number of XAPI MT gateway clients supported, which is the maximum number of gateways the MTA can support.

# of retrieval queue clients The maximum number of XAPI MA retrieval queue clients supported.

# of TCP/IP control blocks The maximum number of TCP/IP connections supported.

# of TCP/IP threads The maximum number of MTA DMOD threads processing TCP/IP connections.

# of TP4 control blocks Maximum number of supported concurrent TP4 connections.

# of TP4 threads Maximum number of MTA DMOD threads processingTP4 connections, including multiple thread connections.

If Internet Mail Service is already installed, the following options will also appear.

Option Description

# of inbound threads The number of threads available to perform content conversion and move inbound mail from the Internet Mail Service to the information store.

# of outbound threads The number of threads available to perform content conversion and move outbound mail to the Internet Mail Service from the information store.

# of InOut threads The number of threads available to perform content conversion and move both inbound and outbound mail between the Internet Mail Service and the information store.

# of threads per processor The number of threads available for traffic between the Internet Mail Service and the Internet.

Option	Description
# of information store buffers	The maximum number of 4KB buffers allocated to the information store database.
# of directory buffers	The maximum number of 4KB buffers allocated to the directory store database.

Option	Description
Minimum # of information store threads	The minimum number of threads that the information store will use to service Messaging Application Programming Interface (MAPI) clients, such as Microsoft Outlook.
Maximum # of information store threads	The maximum number of threads that the information store will use to service MAPI clients, such as Microsoft Outlook.
# of directory threads	The maximum number of threads used by the directory.
Maximum # of concurrent read threads	The maximum number of directory threads available to service replication requests.
# of background threads	The number of threads available for background tasks and to the Gateway In/Out, Send, and Delivery thread pool.
# of heaps	The number of areas of memory used for dynamic memory allocation.
# of private information store send threads	The number of threads that process messages submitted by MAPI clients, such as Microsoft Outlook.
# of private information store delivery threads	The number of threads that the private information store can use to deliver messages to mailboxes.
# of public information store send threads	The number of threads that process public folder replication messages and messages generated by public folder rules.
# of public information store delivery threads	The number of threads that the public information store can use to deliver messages to public folders.
# of information store gateway in threads	The number of information store threads that deliver mail from the message transfer agent (MTA) to the information store for routing elsewhere. For optimal performance, the number of threads should be increased only if the server has multiple processors.
# of information store gateway out threads	The number of information store threads delivering mail from the information store to local mailboxes, or to the MTA. For optimal performance, the number of threads should be increased only if the server has multiple processors.
Buffer Threshold Low Percent	The percentage of available buffers remaining before buffers are flushed to disk. Lower values limit the number of writes to disk; however, low values can degrade performance.
Buffer Threshold High Percent	The percentage of available buffers that must be reached before flushing of buffers to disk stops. For optimal performance, set the percentage value equal to, or slightly greater than, the Buffer Threshold Low Percent.
Maximum # of pool threads	The maximum number of threads servicing Internet connections such as Internet Message Access Protocol, Version 4rev1 (IMAP4), Post Office Protocol version 3 (POP3), and Network News Transfer Protocol ( NNTP) into the information store. This value is per processor.
# of information store users	The number of users for which this server routes mail.
# of concurrent connections to LAN-MTAs	The number of concurrent network associations to MTAs.
# of concurrent connections to RAS LAN-MTAs	The minimum number of concurrent network associations to Remote Access Service (RAS) MTAs.
# of LAN-MTAs	The minimum number of network MTAs supported by the MTA.
# of X.400 gateways	The maximum number of remote MTAs connecting through X.400 using X.25, Transport Control Protocol/Internet Protocol (TCP/IP), or Transport Class 4 (TP4).
ds_read cache latency (secs)	The number of seconds before the directory service read cache expires. Performance improves when items are loaded in the cache by limiting the number of directory service reads.
# of dispatcher threads	The total number of threads used to route messages.
# of transfer threads	The total number of threads used to transfer messages.
# of kernel threads	The total number of threads allocated to process the Open Systems Interconnection (OSI) protocol stack.
# of submit/deliver threads	The total number of MTA submit and deliver threads. For example, if this value is set to 3, the total number of threads is 6. Submit threads receive messages from the information store and deliver threads deliver messages to the information store.
# of RAS LAN-MTAs	The maximum supported number of concurrent network associations to other RAS MTAs.
# of database data buffers per object	The number of 4 KB buffers configured per cached MTA database file. The MTA saves a copy of each message until the message has been accepted by the information store, or another MTA.
# of RTS threads	The total number of threads available to the Reliable Transfer Service (RTS) level of the OSI protocol stack.
# of concurrent MDB/delivery queue clients	The maximum number of information store and XAPI MA delivery queue clients supported. Each client can have more than one session. For optimal performance, the value should be at least 2, to support both the public and private databases.
# of concurrent XAPI sessions	The maximum number of sessions to information store and XAPI MA delivery queue clients, XAPI MA retrieval queue clients, and XAPI MT gateway clients.
Max # of RPC calls outstanding	The maximum number of concurrent remote procedure call (RPC) threads. This limits the maximum number of RPCs that will be processed at one time.
Min # of RPC threads	The minimum number of concurrent RPC threads. This sets the minimum number of RPCs that will be processed at one time.
# of MT gateway clients	The maximum number of XAPI MT gateway clients supported, which is the maximum number of gateways the MTA can support.
# of retrieval queue clients	The maximum number of XAPI MA retrieval queue clients supported.
# of TCP/IP control blocks	The maximum number of TCP/IP connections supported.
# of TCP/IP threads	The maximum number of MTA DMOD threads processing TCP/IP connections.
# of TP4 control blocks	Maximum number of supported concurrent TP4 connections.
# of TP4 threads	Maximum number of MTA DMOD threads processingTP4 connections, including multiple thread connections.

Option	Description
# of inbound threads	The number of threads available to perform content conversion and move inbound mail from the Internet Mail Service to the information store.
# of outbound threads	The number of threads available to perform content conversion and move outbound mail to the Internet Mail Service from the information store.
# of InOut threads	The number of threads available to perform content conversion and move both inbound and outbound mail between the Internet Mail Service and the information store.
# of threads per processor	The number of threads available for traffic between the Internet Mail Service and the Internet.