Windows PerfMon's full name is Windows Performance Monitor. This is a powerful diagnostic tool, which monitors entries including CPU, memory, disk, application, and so on. Here we will focus on storage-related Physical Disk and Processor with CPU performance in the entry, and a closer look at the meaning of each entry.

Using PerfMon:

Open PerfMon is simple. On the Windows desktop, click Start – Run, and type PerfMon. After opening the tools, select Performance Monitor from the left of the options bar. Click the green Add button on the right side of the window, and add entries. Select Physical Disk on Available Counters, and select the disk you want to see on Instance of the Selected Object, then click Add. In the screenshot, we added all of the physical disk performance.

Figure 1

Select the performance indicators and disks which needs monitoring, click OK. PerfMon will start to monitor disk performance data and sampling. PerfMon provides three different ways of presentation: trend charts, histograms, and reports. In fact, image rendering is not important for the Avamar administrator. We need to focus on the data provided by the Average bar on the image below, as shown in Figure 2:

Figure 2

Physical Disk:

From Figure 2, we can see that the properties entry in the Physical Disk contains:

Figure 3

Although the Physical Disk has a large numbers of entries/index involved, but in fact, many of the indicators are calculated using different methods to represent the same concept. Now, we classify these performance entries to Disk Response Time and Disk Queue Length which are common performance indicators. It is much easier to understand and differentiate.

About disk response time entry:

1.     Avg. Disk sec/Transfer: It shows the average time per IO in storage side.

2.     Avg. Disk sec/Read: It shows the average time per read IO in storage side.

3.     Avg. Disk sec/Write: It shows the average time per write IO in storage side.

This display unit is milliseconds. These entries are required first to view content in the performance analysis. Disk Response Time determines the storage system for application service levels, including Avamar. Users often feel a performance issue, but also because the Disk Response Time increased. Following in file systems and database applications to do an example, give some Disk Response Time threshold. If the disk response is in the corresponding range, it is considered acceptable; otherwise you need to see further reason.

File system:

0-10ms: Ideal

0-20ms: An acceptable range.

>20ms: There is a performance problem, require solutions.

Database (Exchange and SQL):

0-10ms: An acceptable range.

>10ms: Need to be optimized.

From the following graph we can see that the Avg. Disk sec/Transfer is 0.004ms:

Figure 4

About Disk Queue Length entry:

1.     Avg. Disk Queue Length: It shows the current disk queue length that the number of how many IO is waiting for the storage.

2.     Avg. Disk Read Queue Length: It shows currently the number of requests waiting to be read by storage.

3.     Avg. Disk Write Queue Length: It shows currently the number of requests waiting to be write by storage.

These values show the information about the disk queue length. Disk Queue is the number of requests that is sent from server side and is waiting to be stored. This entry is also highly correlated with Avamar backup performance. For example, Avamar makes a read request, but the destination disk is working on other tasks. So the new request will be put on the disk queue. Disk queue value is 1. In theory, the Avg. Disk Queue Length value should not be greater than 1. If you see the average Queue Length is greater than 1 during the sampling period, it indicates that storage cannot fully respond to the IO request of Avamar (application side) in sampling period, which means that during this time, Avamar cannot create a backup or backup is very slow.

Other featured entries you may need to focus on:

% idle time: this entry accurately records how much time it remains in the idle state. If this value is below 20%, indicating that the disk system is extremely busy (saturated) state. If the disk system is saturated long-term, you may consider using a faster disk system to replace the existing one.

Processor:

We use the same method, add the Processor performance entry to monitor, as shown:

Figure 5

In many entries, % Processor Time and Processor Queue Length may affect Avamar backup performance.

1.     % Processor Time: It shows the percentage of time that the processor executes a non-idle thread. If this value is greater than 85, the current processor is overused. This situation could lead to avtar (The backup and recovery processes of Avamar) being unable to allocate enough CPU resources to perform the backup job.

2.     Processor Queue Length: It shows the current number of threads in the processor queue. If, within a period of time, the value is twice the CPU, then the CPU does not have enough processing power. In this case, the host may need a faster processor.