Windows 7 / Getting Started

---

Key Performance Monitor Counters

Each and every server, not just your Exchange servers, has key metrics that you will be interested in. These include the following:

• Memory usage and historical memory utilization
• Processor usage and historical processor utilization
• Disk usage and historical disk utilization

Specifically for Exchange servers, you are also interested in metrics covering these additional items:

• Active Directory performance
• Network performance
• MAPI performance

Unsurprisingly, Windows Server provides us with a mechanism for accessing this information in several ways:

• The Performance Monitor tool
• The Resource Monitor tool
• The Task Manager tool
• The .NET Framework API
• The Win32 API

The Win32 API and the .NET Framework API are developer-oriented mechanisms and beyond the scope of our discussion. We will cover the other mechanisms in this section.

Types of Monitoring

Monitoring falls into two broad categories:

Real-Time or Instantaneous In this case, monitoring information is recorded as an event occurs. For example, when a connection is made to an Edge server to begin the process of transferring an email, an entry is written to a log file. This happens in real time, as each event occurs.

Scheduled In this case, a monitoring ''probe'' is executed on a regular basis. For example, every 15 minutes you may check the amount of disk space available on the volume hosting a mailbox store.

Both types of monitoring are valuable. Both can produce immediate "This situation needs to be handled right now!" types of alerts. Both can be used to produce historical trendingbreak information.

Real-time monitoring is often used for producing volume reports (for example, how many email messages were received per hour during each day of the preceding month). Real-time monitoring is also often used for diagnosing failures of one type or another (for example, why a message transfer failed between Server A and Server B).

Scheduled monitoring is generally used for probing specific values of a property against a known optimum value. For example, you can specify that you want an alert to be generated if disk space on a volume goes below 10 percent available, if processor utilization over the last five minutes has exceeded 95 percent, or if Average RPC Latency exceeds 50 ms.

If you think about it, it is easy to see that monitoring is trying to tell us two things about our servers. First, is our server performing at an acceptable level? Second, is our server healthy?

In this case, a healthy server is one that has all of its tested properties fall within ranges that are defined as being acceptable. System Center Operations Manager (OpsMgr) builds hierarchical health for modules on a server (for example, Exchange Server may be one module on a server and the disk may be another, and Exchange Server's health may depend on the disk module being healthy).

Also, you can see that some monitoring measurements may mean nothing by themselves, and become important only when viewed over a period of time to see how they have grown or shrunk. This process is called "trending" and is an important part of server monitoring.

Different roles of Exchange Server will have different optimum values for some things. However, certain considerations apply to all Windows servers, be they Exchange, Active Directory, SQL, File/Print, or (heaven forbid!) some combination of these. The servers require the following:

• Sufficient processor power
• Sufficient memory
• Sufficient disk space

You can consider those to be the "big three" for any server of any type (be it Windows, Linux, Sun, IBM mainframe, or whatever). Most, but certainly not all, performance problems will be traced to a deficiency of one of those items or an improper configuration of software. Any Windows server running Exchange Server will also require the following:

• Sufficient network bandwidth
• Sufficiently fast disk access
• Sufficiently fast Active Directory response
• Properly implemented Active Directory
• Properly implemented DNS

Now that you know the types of monitoring, let's discuss the detailed performance counters and Windows Management Instrumentation (WMI) classes that are relevant for implementing this type of monitoring.