Windows Memory - OS Test

To understand the metrics reported by this test, it is essential to understand how memory is handled by the operating system. On any Windows system, memory is partitioned into a part that is available for user processes, and another that is available to the OS kernel. The kernel memory area is divided into several parts, with the two major parts (called "pools") being a nonpaged pool and a paged pool. The nonpaged pool is a section of memory that cannot, under any circumstances, be paged to disk. The paged pool is a section of memory that can be paged to disk. (Just being stored in the paged pool doesn't necessarily mean that something has been paged to disk. It just means that it has either been paged to disk or it could be paged to disk.) Sandwiched directly in between the nonpaged and paged pools (although technically part of the nonpaged pool) is a section of memory called the "System Page Table Entries," or "System PTEs." The Windows Memory - OS test tracks critical metrics corresponding to the System PTEs and the pool areas of kernel memory of a desktop.

Target of the test : An Amazon Cloud Desktop Group

Agent deploying the test : A remote agent

Outputs of the test : One set of results for every user_on_poweredondesktop

Configurable parameters for the test
Parameter Description

Test Period

How often should the test be executed.

Host

The nick name of the Amazon Cloud Desktop Group component for which this test is to be configured.

Port

Refers to the port at which the specified host listens to. By default, this is NULL.

Inside View Using

To obtain the 'inside view' of performance of the desktops - i.e., to measure the internal performance of the cloud-based virtual desktops - this test uses a light-weight eG VM Agent software deployed on each of the desktops. Accordingly, this parameter is by default set to eG VM Agent.

Report Powered OS

If this flag is set to Yes (which is the default setting), then the 'inside view' tests will report measures for even those desktops that do not have any users logged in currently. Such desktops will be identified by their name and not by the username_on_desktopname. On the other hand, if this flag is set to No, then this test will not report measures for those desktops to which no users are logged in currently.  

Report By User

This flag is set to Yes by default. The value of this flag cannot be changed. This implies that the cloud-based virtual desktops in environments will always be identified using the login name of the user. In other words, in cloud environments, this test will, by default, report measures for every username_on_desktopname.

IsCloudVMs

Since this test runs for a 'Amazon Cloud Desktop Group' component, this flag is set to Yes, by default.

Measurements made by the test
Measurement Description Measurement Unit Interpretation

Free entries in system page table

Indicates the number of page table entries not currently in use by the guest.

Number

The maximum number of System PTEs that a server can have is set when the server boots. In heavily-used servers, you can run out of system PTEs. You can use the registry to increase the number of system PTEs, but that encroaches into the paged pool area, and you could run out of paged pool memory. Running out of either one is bad, and the goal should be to tune your server so that you run out of both at the exact same time. Typically, the value of this metric should be above 3000.

Page read rate in system

Indicates the average number of times per second the disk was read to resolve hard fault paging.

Reads/Sec

 

Page write rate in system

Indicates the average number of times per second the pages are written to disk to free up the physical memory.

Writes/Sec

 

Page input rate in system

Indicates the number of times per second that a process needed to access a piece of memory that was not in its working set, meaning that the guest had to retrieve it from the page file.

Pages/Sec

 

Page output rate in system

Indicates the number of times per second the guest decided to trim a process's working set by writing some memory to disk in order to free up physical memory for another process.

Pages/Sec

This value is a critical measure of the memory utilization on a guest. If
this value never increases, then there is sufficient memory in the guest. Instantaneous spikes of this value are acceptable, but if the value itself starts to rise over time or with load, it implies that there is a memory shortage on the guest.

Memory pool non-paged data in system

Indicates the total size of the kernel memory nonpaged pool.

MB

The kernel memory nonpage pool is an area of guest memory (that is, memory used by the guest operating system) for kernel objects that cannot be written to disk, but must remain in memory as long as the objects are allocated. Typically, there should be no more than 100 MB of non-paged pool memory being used.

Memory pool paged data in system

Indicates the total size of the Paged Pool.

MB

If the Paged Pool starts to run out of space (when it's 80% full by default), the guest will automatically take some memory away from the System File Cache and give it to the Paged Pool. This makes the System File Cache smaller. However, the system file cache is critical, and so it will never reach zero. Hence, a significant increase in the paged pool size is a problem. This metric is a useful indicator of memory leaks in a guest. A memory leak occurs when the guest allocates more memory to a process than the process gives back to the pool. Any time of process can cause a memory leak. If the amount of paged pool data keeps increasing even though the workload on the guest remains constant, it is an indicator of a memory leak.

Committed memory in use in system

Indicates the committed bytes as a percentage of the Commit Limit.

 

Percent

Whenever this measure exceeds 80-90%, application requests to allocate memory in the virtual memory (page file). This ratio can be reduced by increasing the Physical memory or the Page file.

Pool nonpaged failures in system

Indicates the number of times allocations have failed from non paged pool.

Number

Generally, a non-zero value indicates a shortage of physical memory.

Pool paged failures in system

Indicates the number of times allocations have failed from paged pool.

Number

A non-zero value indicates a shortage of physical memory.

Copy read hits in system

Indicates what percent of read I/O being served is coming from system cache, not disk.

Percentage

This is an important counter for applications like the Citrix Provisioning server that stream large volumes of data. If the RAM cache of the server is not sufficiently large, a lot of the I/O requests will be served from the disk, and not the system cache. This will reduce performance. Hence, it is critical to monitor this metric. The higher the value, the better the performance you can see from the server.

Copy reads in system

Indicates how many hits you are really getting.

Reads/Sec

A copy read is a file read operation that is satisfied by a memory copy from a page in the cache to the application's buffer. The LAN redirector uses this method for retrieving information from the cache, as does the LAN server for small transfers. This method is also used by the disk file systems.

Page fault rate in system

Indicates the committed bytes as a percentage of the Commit Limit.

Percent

Whenever this measure exceeds 80-90%, application requests to allocate memory in the virtual memory (page file). This ratio can be reduced by increasing the Physical memory or the Page file.