Exchange 2003 Performance Counters
This is a fairly complete set of useful counters for monitoring Exchange Server performance. According to some of the official monitoring and troubleshooting documentation, there are thresholds for some of these values, and I've included them. Now I will say, it's relatively easy to come up with the counters you are going to use for monitoring, and easy to collect the data, it's more difficult to actually work with the data once you've got it. Unless you have a 3rd-party product to process your counters, there's a tedious manual process involved in averaging values in regular time windows. Not fun. And also remember that, depending on the collection interval you set, your averaging per snapshot will be degraded, and that has to be taken into account.

Additionally, not all of these counters are equal. Some are always critical, some are only referred to when troubleshooting a particular kind of degradation. I've put an asterix next to some critical ones, but my choices aren't definitive.

Logical Disk
* % Free Space < 10%

Memory
Pages/Sec > 1000
Available Bytes < 50mb

MSExchangeIS
* RPC Average Latency > 50 ms at all times
Connection Count (factor 10min inactivity timeout)
Client: Total Reported Latency
Active Connection Count (1 to 3 per real user)
User Count
RPC Requests > 30
Client: RPCs Failed: Server Unavailable
Client: RPCs Failed: Server Too Busy / sec
Client: RPCs Failed: Call Failed / sec
Client: RPCs Failed: Call Cancelled / sec
Client: RPCs Failed: All other errors / sec
Client: RPCs Failed: Access Denied / sec
Client: RPCs Failed / sec
Client: Background RPCs Failed / sec
Client: Foreground RPCs Failed / sec
* Client: Latency > 10 sec RPCs
Client: Latency > 5 sec RPCs
Client: RPCs Succeeded / sec
* RPC Operations / sec .20 per second, per user

MS Exchange IS Mailbox
Peak Client Logons
Active Client Logons > 4 per user
* Average Delivery Time

Paging File
% Usage > 50%

Physical Disk
* Avg. Disk Queue Length < 3 or 2 depending on log/data
Current Disk Queue Length < 3 or 2 depending on log/data
* Avg. Disk Write Queue Length < 3 or 2 depending on log/data
Disk Reads / sec - depends on disk vendor stats
Disk Writes / sec - depends on disk vendor stats

Processor
% Processor Time > 80%

SMTP Server
Remote Queue Length < 1000 & small variance
* Connection Errors / sec
Messages in Local Delivery Queue
DNS Queries / sec
Current Outbound Connections
Current Inbound Connections

System
Processor Queue Length < 2

---------------

These articles go into more depth on how to interpret these counters.

Performance TroubleShooting Basics:
http://msexchangeteam.com/archive/2005/09/28/411674.aspx

RPC Performance Counters Indicate Poor Performance:
http://technet.microsoft.com/en-us/library/d97e8f41-97b8-4858-b51c-bcaf48f48607.aspx

Tuning Exchange 2003 Performance:
http://technet.microsoft.com/en-us/library/d97e8f41-97b8-4858-b51c-bcaf48f48607.aspx

Finetuning Your Exchange Server:
http://gethelp.devx.com/techtips/exo_pro/10min/10min1299.asp

Perf Counters for Users, Logons & Connections:
http://msexchangeteam.com/archive/2006/09/11/428871.aspx  ¶ 9:31 AM