Windows Home Server 2011 : Monitoring Performance (part 5) - Monitoring Performance with Performance Monitor

3. Monitoring Performance with Performance Monitor

For more advanced performance monitoring, Windows Home Server offers the Performance Monitor tool, which you can get to by selecting Start, typing perf, and pressing Enter. In the window that appears, select Monitoring Tools, Performance Monitor, as shown in Figure 8.

Figure 8. Use Performance Monitor to keep an eye on various system settings and components.

Performance Monitor’s job is to provide you with real-time reports on how various system settings and components are performing. Each item is called a counter, and the displayed counters are listed at the bottom of the window. Each counter is assigned a different colored line, and that color corresponds to the colored lines shown in the graph. Note, too, that you can get specific numbers for a counter—the most recent value, the average, the minimum, and the maximum—by clicking a counter and reading the boxes just below the graphs.

The idea is that you should configure Performance Monitor to show the processes you’re interested in (paging file size, free memory, and so on) and then keep Performance Monitor running while you perform your normal chores. By examining the Performance Monitor readouts from time to time, you gain an appreciation of what is typical on your system. Then, if you run into performance problems, you can check Performance Monitor to see whether you’ve run into any bottlenecks or anomalies.

Tip

By default, Performance Monitor samples the performance data every second. To change the sample interval, right-click Performance Monitor and then select Properties. (You can also press Ctrl+Q or click the Properties button in the toolbar.) In the Performance Monitor Properties dialog box, display the General tab, and modify the value in the Sample Automatically Every X Seconds text box. Click OK to put the new sample interval into effect.

Adding Performance Counters

By default, Performance Monitor shows a single counter, % Processor Time, which is the percentage of time the processor is busy. A consistently high value (say, more than 80%) probably indicates a rogue program that needs to be shut down. However, it may also indicate that the Windows Home Server CPU is too slow to keep up with the network demand (although this is unlikely on a home network).

To add another setting to the Performance Monitor window, follow these steps:

1.	Right-click a counter and then click Add Counters. (You can also press Ctrl+I or click the Add button in the toolbar.) The Add Counters dialog box appears, as shown in Figure 9. Figure 9. Use the Add Counters dialog box to add more counters to Performance Monitor.
2.	In the Available Counters list, click the arrow to the right of the counter category you want to work with.
3.	Select the counter you want. If you need more information about the item, activate the Show Description check box.
4.	If the counter has multiple instances, select the one you want from the Instances of Selected Object list. (For example, if you choose Processor as the performance object and your system has either multiple processors or a multiple-core processor, you need to choose which processor you want to monitor. You can also usually select _Total to monitor the total of all the instances.)
5.	Click Add.
6.	Repeat steps 2 through 5 to add any other counters you want to monitor.
7.	Click OK.

Tip

The graph is only useful if you can see the results properly. Unfortunately, sometimes the scale of the graph isn’t appropriate for the numbers generated by a particular counter. The default scale is from 0 to 100, so if a counter regularly generates numbers larger than 100, all you’ll see is a straight line across the top of the graph. Similarly, if a counter regularly generates small numbers, the counter’s graph will be a straight line across the bottom of the graph.

To fix this, you can change the scale used by the Performance Monitor graph. Right-click Performance Monitor and then select Properties. (You can also press Ctrl+Q or click the Properties button in the toolbar.) In the Performance Monitor Properties dialog box, display the Graph tab and modify the values in the Maximum and Minimum text boxes. I also find that activating the Horizontal Grid check box helps you interpret the graph. Click OK to put the new settings into effect.

Understanding Performance Counters

In the Add Counters dialog box, the Available Counters list has dozens of objects, and each object can have dozens of counters. Explaining each one would require another book this size (and would require a level of patience that I don’t have). Fortunately, only a few of the performance objects are truly useful for your Windows Home Server network, and in most situations, you need only track a few counters to monitor the server and network performance. Table 1 presents my list of the most useful performance objects and counters.

Table 1. Useful Performance Monitor Counters for Monitoring Server and Network Performance

Object/Counter	Description
Cache	This performance object represents Windows Home Server’s file system cache, which it uses to hold frequently used bits of data. The more data that Windows Home Server can read from the cache, the faster the system’s performance. See also the Memory\Cache Bytes and Memory\Cache Faults/Sec counters.
Cache\Copy Reads/Sec	This counter tells you the number of times per second that Windows Home Server attempts to locate data in the cache instead of on the disk. Use this counter in conjunction with Copy Read Hits %.
Cache\Copy Read Hits %	This counter tells you the percentage of cache read requests that successfully retrieved data from the cache instead of from the disk. The higher the percentage (anything over 80% is very good), the better the system performance.
Cache\Data Flush Pages/Sec	This counter monitors the number of cache pages that are written back (flushed) to disk per second. If this value is steadily increasing, it might mean that Windows Home Server is having to reduce the size of the cache because memory is getting low.
LogicalDisk/Avg. Disk Queue Length	This value tells you the average number of read and write requests queued for the system’s hard disks during the sample interval. If this value is consistently 2.0 or higher, it probably means that at least one of your hard disks is too slow to keep up with the demand being placed on it. You might want to replace the disk with a faster one.
Memory	This performance object represents Windows Home Server’s memory, which includes both physical RAM and virtual memory.
Memory\Available MBytes	This counter tracks the number of megabytes that are currently available for processes. As this number gets lower, system performance slows because Windows Home Server must reduce the size of the system cache and read more data from the disk. Windows Home Server may also reduce the memory used by services, which can slow performance. If this number drops below 4MB, your system is seriously low on memory. Use the Task Manager to see if a process is using excessive amounts of memory. Otherwise, you may need to add RAM to your system.
Memory\Cache Bytes	This counter tells you the size, in bytes, of the system cache. If the system cache size is falling, it may indicate that Windows Home Server is running low on memory (so it reduces the cache size to free up memory for processes). For content, examine the Memory\Cache Bytes Peak value to see the largest value of the cache size since the system was last booted.
Memory\Cache Faults/sec	This counter monitors the number of times per second that the system looked for data in the system cache but didn’t find it. A steady increase in this value may indicate that the system cache is too small.
Memory\Committed Bytes	This counter measures the number of bytes of physical and virtual memory that the system has committed to running processes. If this value is always close to the value of the Memory\Commit Limit counter—which measures the total amount of physical and virtual memory that can be assigned to processes—it means that either your paging file’s maximum value is too small, or your system doesn’t have enough physical RAM.
Memory\Page Faults/Sec	This counter tells you the average number of page faults that occur per second. This value combines soft page faults and hard page faults.
Memory\Pages Input/Sec	This counter tells you the average number of pages per second that the system is reading to resolve hard page faults. A large number of hard page faults degrades performance because the system must retrieve data from the relatively slow hard disk. You need to either shut down some running programs or services or add RAM. Note, too, that the difference between this value and the Memory\Page Faults/Sec value tells you the number of soft page faults per second.
Memory\Page Reads/Sec	This counter monitors the number of read operations per second that the system is performing to resolve hard page faults. This doesn’t tell you all that much by itself. However, if you divide the Memory\Pages Input/Sec value by Memory\Page Reads/Sec, you learn how many pages the system is retrieving per read operation. A large number of pages per read operation is a sign that your system is low on physical memory.
Memory\Pages Output/Sec	This counter tells you the number of times per second the system writes data to the disk to free up memory. If this value is increasing, your system doesn’t have enough physical RAM.
Memory\Pages/Sec	This value shows the number of pages per second that are retrieved from or written to disk to resolve hard page faults. A consistently large number here (say, more than 2,500 pages per second) probably means that the server doesn’t have enough memory. This counter shows the total for all your hard disks. To narrow down which disk is causing the problem, display separate counters for each instance of the Avg. Disk Queue Length value, as described in the previous steps. (In this case, each instance is a separate hard disk on the server.)
Memory\Pool Nonpaged Bytes	This counter tracks the number of bytes allocated to the nonpaged pool.
Memory\Pool Paged Bytes	This counter tracks the number of bytes allocated to the paged pool, the system memory area that Windows Home Server uses for objects that can be written back to the disk when the system doesn’t need them. (The current size of the paged pool is given by the Memory\Pool Paged Resident Bytes value.) The nonpaged pool and paged pool take memory away from other processes, so if these values are large relative to the total amount of physical memory, you should add more RAM to the system.
Network Interface	This performance object represents Windows Home Server’s network adapter and its connection to the network. For the object instances, select the network adapter you want to monitor (if your system has more than one).
Network Interface\Current Bandwidth	This counter tells you the current network bandwidth, in bits per second.
Network Interface\Bytes Total/Sec	This counter tells you the total number of bytes received and bytes sent over the network connection per second. (This is the sum of the Network Interface\Bytes Received/Sec and Network Interface\Bytes Sent/Sec values.) Multiply this value by 1,024 to calculate the number of bits per second that are passing through the adapter. Under load (say, while streaming media), the result should be close to the Network Interface\Current Bandwidth value. If it’s substantially less, you have a network bottleneck.
Paging File	This performance object represents Windows Home Server’s paging file.
Paging File\% Usage	This counter tracks the current size of the paging file as a percentage of the maximum paging file size. If this value is consistently high—say, 70 percent or more—you either need to increase the maximum size of the paging file, or you need to add more RAM to the system.
Paging File\% Usage Peak	This counter tells you the maximum size of the Paging File\% Usage value in the current session.
PhysicalDisk	This performance object represents Windows Home Server’s hard disks. For the object instances, you can monitor individual hard disks or all the hard disks combined. See also the System\Processor Queue Length counter.
PhysicalDisk\% Disk Time	This counter tracks the percentage of the sample interval that the disk spent processing read and write requests. On your home network, this value should be quite small (usually less than 1%). If you see a larger value, you may have a hard disk that’s too slow.
Processor	This performance object represents Windows Home Server’s CPU. If your system has multiple processors or a multiple-core processor, you can select an individual processor or core as an instance.
Processor\% Idle Time	This counter tells you the percentage of time during the sample interval that the processor was idle.
Processor\% Interrupt Time	This counter shows the percentage of time during the sample interval that the processor was processing interrupt requests from devices.
Processor\% Privileged Time	This counter tells you the percentage of time during the sample interval that the processor spent running code in privileged mode, a processing mode that gives operating system programs and services full access to system hardware.
Processor\% User Time	This counter tells you the percentage of time during the sample interval that the processor spent running code in user mode. On your Windows Home Server network, this value—as well as the values for Processor\% Idle Time, Processor\% Interrupt Time, and Processor\% Privileged Time—should be at or near 0 most of the time. If any one of these values is consistently high, you might need to upgrade to a faster processor or a processor with more cores, or add a second processor if your system motherboard supports this.
System	This object represents the Windows Home Server system as a whole.
System\Processor Queue Length	This counter tells you the number of threads that are waiting to be executed by the processor (or processors; there is just one queue for all CPUs). If this value is consistently 10 or more, your processor isn’t doing its job, and you should consider upgrading it or adding a second processor (if possible).
System\System Up Time	This counter shows the time, in seconds, that Windows Home Server has been running since the most recent boot.