Have you ever wondered what's going on inside your server? Performance monitoring is vital to keeping your server running smoothly over time. If you keep an eye on the performance of your system’s components (such as memory and the processor), you'll know when you need to upgrade them—before they create problems or cause bottlenecks. But finding bottlenecks is only one reason for monitoring your system’s performance. If you know the ranges of your counters when your system is healthy, you’ll be able to spot problems when your server starts to experience performance-related problems. In this Daily Drill Down, I'll tell you what to look for when you monitor the performance of a Windows 2000 server.
The System Monitor
If you're a Windows NT user, you're probably familiar with the Performance Monitor. However, Performance Monitor no longer exists in Windows 2000. A new tool called System Monitor has replaced it. Like almost every other tool under Windows 2000, the System Monitor is actually a Microsoft Management Console snap-in. Since the System Monitor is based on Microsoft Management Console, you can create custom snap-ins that contain those counters that you need for various situations. System Monitor offers the same basic functionality that Performance Monitor did; however, System Monitor’s capabilities have been extended. For example, with System Monitor, you can modify the fonts, colors, and borders of graphs. You also can use the clipboard to cut and paste counter paths to and from Web pages. And best of all, you can save your counters as an HTML file and post them on the Web or print them for future reference.
You can access the System Monitor by double-clicking the Performance icon in the Control Panel. Afterwards, you'll see the main System Monitor screen, as shown in Figure A.
Before we delve too deeply into performance monitoring, let's take a moment to review the various components within the System Monitor. Just above the graph, you'll see a toolbar. This toolbar will allow you to perform such basic operations as adding and removing counters from the chart, freezing the display, and changing view modes.
The graph, which shows you the results of the various counters, is highly customizable. Simply right-click the graph, select the Properties command from the resulting context menu, and you'll see the graph's properties sheet. This sheet lets you customize all sorts of things, including the style, colors, and fonts of the graph.
Just below the graph is the value bar. It indicates the last, average, minimum, and maximum values of a selected counter. Finally, below the value bar is the legend. The legend defines the performance aspect of a particular color on the graph.
Using the System Monitor
Now, let me explain how to use the System Monitor to check out your server's processor utilization. Begin by clicking the Plus icon on the toolbar. You'll see the Add Counters dialog box, as shown in Figure B. This dialog box is divided into several sections. First, you can select which computer you want to monitor. For this example, let’s monitor the local computer. The Performance Object drop-down menu allows you to choose which general aspect of the system you want to measure. For example, you can look at the processor, memory, or print queues—just to name a few. Below that are the actual counters, which are related to the performance object that you selected. You can monitor all of the available counters (but I don’t recommend doing so), or you can select a specific counter from the list. Let’s stick with the Processor Performance Object and look at the % Processor Time counter. Click Add to add the % Processor Time value to the graph. But first, notice that Figure B contains an explanation of the % Processor Time counter. We accessed this explanation by selecting the counter with which we wanted to work and clicking Explain. You can use this method whenever you're unsure about what a particular counter does. After you've added all of the necessary counters to the graph (in this case, just the % Processor Time counter), click Close to return to the graph.
When you return to the graph, the % Processor Time counter will appear on the graph, as shown in Figure C. The graph indicates what percentage of the server's total processing power is being used at any given moment. Look at the legend, and notice that the red line is associated with the % Processor Time counter. In this case, it’s pretty obvious, but the legend comes in handy when you're dealing with more than one counter. Now, look at the value bar and find out what were the lowest and highest percentages of used processing power. The value bar also indicates the average value of used processing power and the most recent value.
Once you've created a graph, you can control the way that you see it. All you have to do is click the appropriate toolbar icon to view the information as a graph, a histogram, or a report, as shown in Figure D.
The graph itself is highly customizable. You can do pretty much anything that you want to the graph by right-clicking it. When you do, you'll see a context menu with three choices: Add Counters, Save As, and Properties. The Add Counters option lets you add more counters to the graph. The Save As option allows you to export the graph as a static HTML page, which you can post on the Web. You also can pull the HTML page into Internet Explorer or Microsoft Office and print it. The Properties command allows you to control the appearance of the graph. You can control such factors as color, font, title, borders, and other aspects. As you can see in Figure E, you can drastically change the way the graph looks.
Watch out for bottlenecks
Of course, one of the things that System Monitor is really good for is spotting system bottlenecks. However, you can choose from hundreds of counters. So, which of these counters are really important? And even if you know which counters are important, how do you know what indicates a bottleneck? For example, I showed you how to set up the % Processor Time counter. If it were one of the more important counters, you may wonder what percentage of processing time is too high. Let’s review some of the more important counters and examine the recommended threshold for those counters. Keep in mind that just because we don't discuss a particular counter doesn't mean that it won’t come in handy for you. These are merely general counters that should be monitored on most Windows 2000 servers. You can set up Windows 2000 in such a way that it will take action when a counter exceeds a threshold that you recommend. First, let's look at the appropriate values for important counters.
You'll definitely want to keep an eye on your system's hard disk. There are several counters that you'll want to watch so that you can see how your hard disk is holding up. In the Logical Disk category, you should keep an eye on the % Free Space and the % Disk Time counters. The % Free Space should never drop below 15 percent. Likewise, the % Disk Time should stay below 90 percent. A high % Logical Disk Time value may indicate excessive paging; your system may need more RAM. You also should keep an eye on the Current Disk Queue Length counter in the Physical Disk section. If this value averages the number of spindles on your system plus three or more, then disk requests are backing up, and you may need more RAM or a faster hard disk.
To keep an eye on your system's memory usage, monitor the Available Bytes and Pages Per Second counters in the Memory category. The Available Bytes counter should never drop below 4 MB. If it does, you should check your system for a memory leak. Fix the leak if one exists. If no memory leak exists, you'll need to add more RAM. The Pages Per Second should never average above 20. A high value indicates the need for more memory.
In the Network category, you should keep an eye on the % Net Utilization counter. This counter relates to the total amount of network bandwidth being consumed. This number will vary depending on the type of network that you're using. For Ethernet networks, however, this number shouldn't average above 30. A high average number may indicate the need to segment your network.
The Page File is a file on your hard disk that acts as physical RAM. Every time the RAM is full, objects are taken out of RAM and placed into the Page File until they’re needed again. That way, your memory will have more room. Under the Page File category, you should keep an eye on the % Usage counter. If this counter averages above 70 percent, you need more RAM. You can confirm this fact by looking at the other two memory counters.
The most important counter that’s related to the processor is the % Processor Time counter. Although it's normal for this counter to spike to 100 percent occasionally, it should never average above 85 percent. A high average indicates the need for a processor upgrade.
If you're running a multi-processor machine, you can see how the processors compare by looking at the Processor Queue Length counter under the System category. This value should stay under three. However, this counter tends to read inaccurately. Be sure to look at the average value over a long period of time before you decide to correct the problem by adding more or faster processors.
In the Server category, check the Bytes Total/Sec counter. You should look at this value on all of your servers and add the results together. If the value approaches the total transfer rate for your network, then you need to segment the network.
In this Daily Drill Down, I explained the importance of keeping an eye on your server's performance, and I introduced you to the System Monitor tool. After showing you how to use System Monitor, I discussed some of the counters. In the next part of this series, I'll tell you how to take advantage of the logging and alerting capabilities that System Monitor offers.
Brien M. Posey is an MCSE who works as a freelance technical writer and as a network engineer for the Department of Defense. If you’d like to contact Brien, send him an e-mail. (Because of the large volume of e-mail he receives, it's impossible for him to respond to every message. However, he does read them all.)The authors and editors have taken care in preparation of the content contained herein, but make no expressed or implied warranty of any kind and assume no responsibility for errors or omissions. No liability is assumed for any damages. Always have a verified backup before making any changes.