Exchange 2003 - Important Basics for a Smooth Running System

• Use Good Hardware
• Use Hardware on the Hardware Compatability List
• Use ECC Memory
• Give the server plenty of memory for ts requirements
• Use Hardware RAID wherever possible:

C Drive Mirrored,

SMTP Queues Striped and Mirrored

Transaction Logs Striped and Mirrored

Database Files (either stripe and mirror or RAID 5)

• Keep Logs and Database seperate where possible
• Have at least 110 percent disk space spare to running ISINTEG/ESEUTIL
• Do not use disk compression
• Set Transaction Logs to overwrite and max 16Mb size
• Set Dr Watson as default debugger (usually standard, but may have been changed by another program)
• Ensure you have more than one DC
• If you have Exchange Corporate Edition/ Server 2003, and have more than 1GB RAM, you must set the 3GB switch ON
• Set the recovery options
• Disable Circular Logging
• Set Maximum Mailbox quotas to ensure files do not grow bigger than they should
• Set Group Permissions rather than User Permissions
• Do full online Backups and Check the restores work correctly
• Full Online backups purge old logs after doing a checksum verification
• Check your disaster recovery plans actually work
• Go through event logs and fix problems as they occur
• Check backups go through OK.
• Check for updates
• Backup the whole server
• Use the Baseline to actually compare what the mail is doing in terms of flow and speed
• Do an ESEUTIL defrag every 12 months or after a large data move

Use PerfMon to see if bottlenecks are developing:

PerfMon Values:

Database \Log Record Stalls/sec:

Average value should be less than 10 per sec

There should be no spikes higher than 100 per second

This indicates the number of log records that cannot be added to the buffer per second because the log buffers are already full. Buffers are set to : Exchange 2000 this is 84 and Exchange 2003 this is 500.

Database\Log Threads Waiting

Should be below 10 for an average value

Indicates the number of threads waiting to complete an update of the database by writing their data to the log. If this is too high, there may be a bottleneck

MSExchange IS\RPC Requests:

Should be below 30 at all times

Indicates number of MAPI RPC requests presently being serviced by the Information Store service. Information Store can only service 100 RPC requests before rejecting client requests

MSExchangeIS\Average Latency:

Should be below 50ms at all times

Indicates the RPC latency in milliseconds, averaged for last 1024 packets. Usually 10-20ms on healthy servers.

MSExchangeIS\RPC Operations per sec:

This indicates how many RPC operations are being asked of the Exchange store per second and how many it is actually responding to per second. Should rise and fall in synch with RPC requests. Flatlining is very dangerous

MSExchangeIS\\VirusScan Queue Length

Current number of outstanding requests that are queued for virus scanning

Active Client Logons

This is the number of clients that performed any action within the last 10 minutes

Paging File\% Usage

Should be below 50%

A high value indicates you need to increase pagefile size or get more RAM

Memory\Available MBytes

There should be 50Mb of available memory at all times

This counter indicates the amount of physical memory immediately available for allocation to a process   or for system use.

Memory\Pages/sec

Should be below 1000 at all times

This counter is a primary indicator of the types of faults that cause system wide delays

Memory\Pool NonPaged Bytes

There must be no more than 100Mb of non paged pool memory being used

The kernel non page memory pool is an area of system memory (physical memory being used by the system) for kernel objects that cannot be written to disk but must remain in memory while the objects are allocated.

Memory\Pool Paged Bytes

Unless a backup or restore is taking place, there must not be more than 180Mb of paged pool memory being used.

Indicates the number of bytes in the kernel memory paged pool

IF YOU HAVE A SAN:

Physical Disk\Average Disk sec/Read

Indicates average time in secs to read data from the disk.

1. Database drive:

Average value should be below 20ms

Spikes no higher than 100ms

2. Log Drive:

average<5Ms

Spikes not higher than 50ms

3. SMTP drive

Average<10ms

Spikes not higher than 50ms

Physical Disk\Average Disk sec/Writes

Indicates average time in secs to write data from the disk.

1. Database drive:

Average value should be below 20ms

Spikes no higher than 100ms

2. Log Drive:

average<10Ms

Spikes not higher than 50ms

3. SMTP drive

Average<10ms

Spikes not higher than 50ms