PerfMon provides server-wide real-time
and logged performance monitoring. First introduced with Windows NT
4.0, the core features and user interface have barely changed from the
first Microsoft Management Console (MMC) snap-in. In Windows Server
2003 the tool was renamed to System Monitor, although the data logging
functionality of System Monitor retained the name Performance Monitor.
In Windows Server 2008, PerfMon was incorporated into Reliability and
Performance Monitor.
You can use Performance Monitor for many common tasks:
- View real-time performance data on your server.
- See performance data represented visually.
- Record performance data over an extended time frame.
- Quantify the performance impact of hardware or software changes.
- Save and export performance data.
- Fire alerts based on performance thresholds.
- Compare performance data from different servers.
- Capture a baseline set of performance counters for trending and troubleshooting over time.
Reliability and Performance Monitor
PerfMon in Windows Server 2008 brings a
new look and a new name for the parent snap-in, Reliability and
Performance Monitor, although real-time performance monitoring retains
the PerfMon name.
Reliability and Performance Monitor comprises
three components: Monitoring Tools, Data Collector Sets, and Reports.
Resource Overview
Once Reliability and Performance
Monitor is launched, the Resource Overview screen is displayed showing
real-time performance data. The Resource Overview provides a visual
representation of each of the four key hardware elements: CPU, Disk,
Network, and Memory. Each element can be expanded to reveal a list of
processes, listed in descending order by resource type; for example,
when CPU is expanded, all processes are listed ordered by Average CPU
descending, as shown in Figure 1.
In addition to the four individual performance
graphs displayed across the top of the Resource Overview, each resource
element displays two mini-performance bars. It’s worth noting that for
CPU, when looking at this on your own PC, the green bar displays the
current CPU utilization and the blue bar displays the maximum CPU
frequency. Some computers may have maximum CPU frequency less than 100%
when operating in energy-saving modes. I
Data Collector Sets
Data Collector Sets combine all the
information necessary for common problem diagnostics, including event
tracing, performance counters, and configuration (Registry and WMI
classes). Administrators can create Data Collector Sets with providers
enabled for trace and counter data. Once a collector set has been
defined, it is stored in Reliability and Performance Monitor. This
enables starting and stopping the collector at any point in the future
without recreating it, or it can be controlled on a schedule.
Three pre-defined system Data Collector Sets are
included: LAN Diagnostics, System Diagnostics, and System Performance.
Each collects performance counters, trace data, and system
configuration for common troubleshooting scenarios.
Reliability Monitor
Reliability Monitor provides a system
stability chart. Here, events such as hardware failures, or application
or Windows failures, are tracked against a timeline. The data presented
by Reliability Monitor provides access to failure activity information,
plotted against a time chart to facilitate correlation between failure
events and system activity (software installation or uninstallation,
etc.).
The chart displayed by Reliability Monitor plots the System Stability Index,
a rating system reflecting reliability where 10 is a stable server and
1 is considered an unstable server. The purpose of the System Stability
Index is to assist in correlating a decrease in system stability with a
specific change (such as a new device driver or a hotfix installation).
PerfMon Usability in Windows Server 2008
There are some user interface
enhancements for PerfMon in Windows Server 2008 (compared with Windows
Server 2003 and earlier). There are relatively minor changes between
Windows Server 2008 and the R2 release; this section contains a summary
of the highlights.
Auto-Scaling Counters
In early versions of Windows Server,
counter values were often either off the top of the PerfMon graph or
dwarfed by other counters — neither situation enabled users to easily
see changes in these values. This made scaling counters a painful
process of trial and error, as each counter had to be selected in turn
in an attempt to choose a reasonable value to scale the counter by. The
process was made much simpler in Windows Server 2008 because users
could select a group of counters, right-click, and choose “Scale
selected counters.” Windows then adjusted the scale of each counter to
a reasonable value so that all lines are plotted in or around the
middle of the graph.
NOTE Always
be aware of auto-scaling in PerfMon. Check the scaling of counters
before comparing multiple counters, particularly when comparing between
servers. Auto-scaling can adjust instances of the same counter to use
different scales.
Show/Hide Counters
Another minor but useful enhancement to
PerfMon in Windows Server 2008 enabled the showing or hiding of
counters on the graph. This is useful when monitoring in real time
because many counters can be captured and fewer can be displayed.
Showing and hiding counters means the data (including min, max, and
average values) is still recorded but it can be displayed as needed,
which is also faster than adding counters.
