WINDOWS 1002000 PROFESSIONAL RESOURCE KIT [Electronic resources]

Chris Aschauer

نسخه متنی -صفحه : 335/ 277
نمايش فراداده

Establishing a Baseline for Processor Performance

Begin your monitoring routine with an examination of processor usage under your normal workload. By doing so, you can begin to establish a baseline or reference point for processor usage. The baseline is generally not a single value, but a range within which processor usage can fluctuate and still provide acceptable performance. You can use the baseline to identify trends, such as increasing processor demands over time, or to recognize problems that arise from a sudden change.

Selecting Counters for Baseline Monitoring

To determine the baseline, use the following counters to create logs of processor usage over an extended period (from several weeks to a month).

    Processor % Processor Time

    SystemProcessor Queue Length

Be aware of the Idle process when monitoring processor usage. The Idle process runs a thread on each processor. This thread runs when the system is not already running the thread of an active user or system process. System Monitor and Task Manager both use the Idle process to calculate time when the processor is not busy. You can see processor time for the Idle process on the Processes tab in Task Manager (called the System Idle Process) or by tracking the Process(Idle) % Processor Time counter in System Monitor. Notice that the Total instance for this counter includes processor time for the Idle process. To measure the Idle process, use the Process(Idle) % Processor Time counter, or use the Processes tab in Task Manager. Zero idle time could mean that the processor is handling a lot of work, but it could also mean that the processor or central processing unit (CPU) is overloaded.

Selecting Times for Baseline Monitoring

To monitor processor activity, log the counters of the System, Processor, Process, Thread, PhysicalDisk, and Memory for at least several days at an update interval ranging from 15 minutes to an hour. (Use much shorter intervals for bottleneck detection.) Include network counters such as Bytes Total/sec (on the Network Interface object) if you suspect that network traffic might be interrupting the processor too frequently. Because excessive demand on memory and disk resources can cause bottlenecks that appear to affect your processor's performance, also include disk and memory counters in your monitoring configuration to help you determine the true source of any processor bottleneck.

If any applications running on the computer have counters, use these counters to monitor their activity and monitor these values along with system counter values.

Track the values reported at various times of day-for example, while users are logging on or off, while backups are being done, and so forth. As you are monitoring values for these counters, you might see occasional spikes. Typically, you can exclude these from your baseline; the range of values that appear consistently are the ones that constitute your baseline.

NOTE

Keep in mind application overhead and disk-space usage when you set your monitoring frequency. Frequent updating demands more work and more file storage capacity from your computer. You can experiment with different update intervals to balance these considerations against the level of detail you require for your monitoring data. For more information about how to monitor performance, see Windows 2000 Professional Help.

The longer you log, the more accurate your baseline will be. Processor use might be a problem only at certain times of the day, week, or month, and you are likely to see patterns in your workload that are correlated with changes in processor activity if you log for a longer duration. You can even use the log service to schedule monitoring at critical times to determine whether your processor is operating efficiently. During these critical times, you might want to log intervals as short as every two seconds to get an accurate picture of processor usage on your system. This helps you isolate those applications that heavily stress your processor for further investigation and monitoring.