Optimizing Windows 7 with Performance Monitor (part 3)

4. Managing Performance Monitor Data with Collector Sets

The data collector set portion of Performance Monitor is shown in Figure 17. Data collector sets are used to collect data into a log so that the data can be reviewed and saved for comparison at a later date (baselining). You can view the log files with Performance Monitor, as described in the previous section, "Managing Performance Monitor Properies" (in the "Source Tab" subsection).

Figure 17. Performance Monitor data collector set configuration

Data collector sets can collect the following data:

• Performance counters

• Event trace data

• System configuration information

There are two built-in data collector sets that track multiple counters for system diagnostics and system performance. You can also create your own user-defined data collector sets and save them for later use. You can view the reports from these data collector sets within the Reports folder in Performance Monitor.

Performance counter logs record data about hardware usage and the activity of system services. You can configure logging to occur manually or on a predefined schedule. Perform Exercise 2 to create a data log.

4.1. Creating an Alert

Alerts can be generated when a specific counter rises above or falls below a specified value. You can configure alerts to log an entry in the application event log and/or start a data collector set. Creating an alert is similar to creating a performance counter data log except you are required to specify the alert conditions. You create an alert by creating a user-defined collector set, manually specifying counters. In the Create New Data Collector Wizard, you have the option of selecting a Performance Counter Alert radio button. For example, you might configure a performance counter alert that will log an entry whenever the % Free Space counter for C: falls below 5 percent. After you create the alert, you can modify the alert parameters by right-clicking the data collector and selecting Properties.

After you create the data collector sets and set the alerts, you will want to run the sets and save the logs periodically. Reviewing the logs gives you a proactive approach to managing your Windows 7 performance.

Now let's take a look at how to manage Windows 7 performance.

5. Managing System Performance

By analyzing data, you can determine whether any resources are placing an excessive load on your computer, resulting in a system slowdown. The following list gives some of the causes of poor system performance:

• A resource is insufficient to handle the load that is being placed upon it, and the component might need to be upgraded, or additional components might be required.

• If a resource has multiple instances, the resources might not be evenly balancing the workload, and the workload might need to be balanced over the multiple instances more effectively.

• A resource might be malfunctioning. In this case, the resource should be repaired or replaced.

• A specific program might be allocating resources improperly or inefficiently, in which case the program needs to be rewritten or another application should be used.

• A resource might be configured improperly and causing excessive resource usage, and need to be reconfigured.

There are four main subsystems that you should monitor. You should configure counters in your data collector set for each of the following:

• The memory subsystem

• Theprocessorsubsystem

• The disk subsystem

• The network subsystem

Each subsystem should be examined over time to evaluate Windows 7 performance.

5.1. Monitoring and Optimizing Memory

When the operating system needs a program or process, the first place it looks is in physical memory. If the required program or process is not in physical memory, the system looks in logical memory (the page file). If the program or process is not in logical memory, the system then must retrieve the program or process from the hard disk. It can take thousands of times longer to access information from the hard disk than to get it from physical RAM. If your computer is using excessive paging, that is an indication that your computer does not have enough physical memory.

Insufficient memory is the most likely cause of system bottlenecks. If you have no idea what is causing a system bottleneck, memory is usually a good place to start checking. To determine how memory is being used, you need to examine the following two areas:

Physical Memory The physical RAM you have installed on your computer. You should have as much memory as either your machine or budget can handle. It's actually a good idea to have more memory than you think you will need just to be on the safe side. As you've probably noticed, each time you add or upgrade applications, you require more system memory.

Page File Logical memory exists on your hard drive. If you are using excessive paging (swapping between the page file and physical RAM) or hard page faults, it's a clear sign that you need to add more memory.

The first step in memory management is determining how much memory your computer has installed and what the appropriate memory requirements are based on the operating system requirements and the applications and services you are running on your computer.

5.2. Key Counters to Track for Memory Management

The following three counters are the three most important counters for monitoring memory:

Memory => Available Mbytes Memory => Available MBytes measures the amount of physical memory that is available to run processes on the computer. If this number is less than 20 percent of your installed memory, it indicates that you might have an overall shortage of physical memory for your computer, or you possibly have an application that is not releasing memory properly. You should consider adding more memory or evaluating application memory usage.

Memory => Pages/Sec Memory => Pages/Sec shows the number of times the requested information was not in memory and had to be retrieved from disk. This counter's value should be below 20; for optimal performance, it should be 4 or 5. If the number is above 20, you should add memory or research paging file use more thoroughly. Sometimes a high Pages/Sec counter is indicative of a program that is using a memory-mapped file.

Paging File => % Usage Paging File => % Usage indicates the percentage of the allocated page file that is currently in use. If this number is consistently over 70 percent, you might need to add more memory or increase the size of the page file. You should track this counter in conjunction with Available MBytes and Pages/Sec.

The counters listed previously work together to show what is happening on your system. Use the Paging File => % Usage counter value in conjunction with the Memory => Available MBytes and Memory => Pages/Sec counters to determine how much paging is occurring on your computer.

Along with memory counters, processor (or CPU) counters are valuable in evaluating Windows 7 performance.

5.3. Managing Processor Performance

Processor bottlenecks can develop when the threads of a process require more processing cycles than are currently available. In this case, the process will wait in a processor queue, and system responsiveness will be slower than if process requests could be immediately served. The most common causes of processor bottlenecks are processor-intensive applications and other subsystem components that generate excessive processor interrupts (for example, disk or network subsystems).

In a workstation environment, processors are usually not the source of bottlenecks. However, you should still monitor this subsystem to make sure that processor utilization is at an efficient level.

Let's now take a look at some of the key counters that you need to track when watching the processor.

5.4. Key Counters to Track for the Processor

You can track processor utilization through the Processor and System objects to determine whether a processor bottleneck exists. The following three counters are the most important counters for monitoring the system processor:

Processor => % Processor Time Processor => % Processor Time measures the time that the processor spends responding to sysrem requests. If this value is consistently above an average of 85 percent, you might have a processor bottleneck. The Processor => % User Time and Processor => % Privileged Time counters combine to show the total % Processor Time counter. You can monitor these counters individually for more detail.

Processor => Interrupts/Sec Processor => Interrupts/Sec shows the average number of hardware interrupts received by the processor each second. If this value is more than 3,000, you might have a problem with a program or hardware that is generating spurious interrupts. (This value will vary in optimization based on the processor type; you'll need to do a little research for your specific processor to see the appropriate value.)

System => Processor Queue Length System => Processor Queue Length is used to determine whether a processor bottleneck is due to high levels of demand for processor time. If a queue of two or more items exists for an extended period of time, a processor bottleneck might be indicated. If you suspect that a processor bottleneck is due to excessive hardware I/O requests, then you should also monitor the System => File Control Bytes/Sec counter.

5.5. Tuning and Upgrading the Processor

If you suspect that you have a processor bottleneck, you can try the following solutions:

• Use applications that are less processor-intensive.

• Upgrade your processor.

• If your computer supports multiple processors, add one.

The Memory and Processor subsystem objects are important counters to evaluate in determining your Windows 7 performance. You should also look at the hard drive or disk subsystem to check for issues as well.

6. Managing the Disk Subsystem

Disk access is the amount of time your disk subsystem takes to retrieve data that is requested by the operating system. The two factors that determine how quickly your disk subsystem will respond to system requests are the average disk access time on your hard drive and the speed of your disk controller.

6.1. Key Counters to Track for the Disk Subsystem

You can monitor the PhysicalDisk object, which is the sum of all logical drives on a single physical drive, or you can monitor the LogicalDisk object, which represents a specific logical disk. The more important counters for monitoring the disk subsystem are as follows:

PhysicalDisk => % Disk Time and LogicalDisk => % Disk Time PhysicalDisk => % Disk Time and LogicalDisk => % Disk Time show the amount of time the disk is busy because it is servicing read or write requests. If your disk is busy more than 90 percent of the time, you will improve performance by adding another disk channel and splitting the disk I/O requests between the channels.

PhysicalDisk => Current Disk Queue Length and LogicalDisk => Current Disk Queue Length PhysicalDisk => Current Disk Queue Length and LogicalDisk => Current Disk Queue Length indicate the number of outstanding disk requests that are waiting to be processed. On average, this value should be less than 2.

LogicalDisk => % Free Space LogicalDisk => % Free Space specifies how much free disk space is available. This counter should indicate at least 15 percent.

6.2. Tuning and Upgrading the Disk Subsystem

When you suspect that you have a disk subsystem bottleneck, the first thing you should check is your memory subsystem. Insufficient physical memory can cause excessive paging, which in turn affects the disk subsystem.

If you do not have a memory problem, try the following solutions to improve disk performance:

• Use faster disks and controllers.

• Confirm that you have the latest drivers for your disk adapters.

• Use disk striping to take advantage of multiple I/O channels.

• Balance heavily used files on multiple I/O channels.

• Add another disk controller for load balancing.

• Use Disk Defragmenter to consolidate files so that disk space and data access are optimized.

After you evaluate the first three subsystems—memory, processor, and disk—you also need to look at the network subsystem to optimize your Windows 7 performance.

7. Optimizing the Network Subsystem

Windows 7 does not have a built-in mechanism for monitoring the entire network. However, you can monitor and optimize the traffic that is generated on your Windows 7 machine. You can monitor the network interface (your network card) and the network protocols that have been installed on your computer.

Network bottlenecks are indicated when network traffic exceeds the capacity that can be supported by the local area network (LAN). Typically, you would monitor this activity on a network-wide basis—for example, with the Network Monitor utility that ships with Windows Server 2003.

7.1. Key Counters to Track for the Network Subsystem

If you are using the Performance Monitor utility to monitor local network traffic, the following two counters are useful for monitoring the network subsystem:

Network Interface => Bytes Total/Sec Network Interface => Bytes Total/Sec measures the total number of bytes sent or received from the network interface and includes all network prorocols.

TCPv4 => Segments/Sec TCPv4 => Segments/Sec measures the number of bytes sent or received from the network interface and includes only the TCPv4 prorocol.

7.2. Tuning and Upgrading the Network Subsystem

You can use the following guidelines to help optimize and minimize network traffic:

• Install and configure only the network protocols you need.

• Use network cards that take advantage of your bus speed.

• Use faster network cards—for example, 100 Mbps Ethernet or 1 Gbps Ethernet instead of 10 Mbps Ethernet.