Loadrunder Scene Design--Add Windows Resource counter and indicator description

Objective

Monitor resource usage for Windows servers to be tested

A. Adding A counter step

1, add view, double-click, drag the left side of the Windows rescources to the right of the graphics area, or right side of the existing view, Open a New Graph, find Windows Rescource, add

2. Right-click already added Windows Resource, select Add Measurements ...

3. In the pop-up window, click Add, pop up add machine adding the IP to monitor the system (usually the target server), select the System platform

Click OK, as follows

If necessary, you can continue to add measurements

Once added, click OK to save

B. Description of the Counter key indicator ( all thresholds are averages. )

a . Processor:

    1 % Processor Time: processor times / Span style= "font-family: Arial" > usage mean ( CPU ) threshold range is: less than 85% average is greater than 85% or 95%, suspected processor bottleneck. Its figure occasionally higher to reach the 100%

2)Processor Queue Length: If this Processor Queue Length remains unchanged (>=n ( number of cores ) +1) and the processor utilization average%processor time exceeds 90%, there is a potential for processor bottlenecks;Processor Queue Length exceeds n+1,processor Time is very low , then perhaps more should be to solve the problem of processor blocking, which is the performance is generally not a bottleneck.

two . Memory

1) memoryavailable MBytes: The current available memory, at least 1% of the physical memory value! If the available memory in the Windows system is less than the total number of physical memory 1%, the memory may be a bottleneck. If its available memory trend drops first and then rises steadily again, indicating normal, representing memory is used, and then released. If the memory is not released, that is, the stability does not rise after the drop, which means that memory is not released, indicating a possible memory leak.

2) when the processor takes data (or code) into memory, it is called a page fault if an error occurs.

A. if all data (or code) is found elsewhere in the memory, it is called a soft error

B. If the page is not found in memory and found on disk, it is called a hard error (pages/sec)

C. the system can run normally without a large number of soft errors . However, if there are a large number of hard errors in the system, it will seriously affect the operation and performance of the system. In general, the hard error (unit: number) threshold is: memory 1%, that is , 2G of memory, hard error not more than three. Soft Error + hard error =page fault

3) page Reads/sec(page read rate): The number of reads from the hard disk in order to resolve a hard error! If Page read/sec continues to be larger than the 1% of physical memory , it means that there is not enough memory!

4) pages/sec refers to the number of pages that resolve hard errors that are read from or written to disk , which is high enough to indicate low memory

Three Systems (System)

①context switches/sec: more times, it will occupy the system resources. The reason is that program code inefficiencies or system structural design flaws!

four . Physical Disk ( Physical Disk )

1) Disk Utilization (%disk time): One of the performance tuning core points: minimize disk IO.

2) Avg.Disk Queue Length: The average number ofdisk read / write requests (queued in the instance interval of the selected disk). The threshold value is: no more than 1.5~2 times the disk. To improve performance, you can increase the disk.

3) Disk Reads(Write)bytes/s: If more than dozens of m or hundreds m , the disk disk bottleneck is suspected.

Five. Network Interface

A) Bytes total/sec: determines the network bandwidth, which can be compared with the current network bandwidth. (1Byte =8bits). The value *8 is compared to half of the bandwidth, and if the value is less than half the bandwidth, it is generally considered that the network has no bottleneck and that the bandwidth is in bits.

Six. Bottleneck Judgment:

(a) to determine the problem of memory leaks: (This situation generally requires a long time to run)

If a memory leak occurs,the values of the Process/private bytes counter and the process/working set counter tend to rise, while The value of the memoryavailable MBytes is reduced.

( two ) determine the application's problems

If the system code is inefficient or the structural design is flawed, resulting in a large number of context switches ( context switches/sec display is too high), then it consumes a lot of system resources. If the throughput of the system is reduced and the CPU usage is high, and this phenomenon occurs when switching levels above 15000 , it means that the context

Seven Experience

A) if the "timeout error" reaches more than 90%, you may need to increase the hardware configuration

If there are more "internal service Errors", you may need to increase the presence of the program

Loadrunder Scene Design--Add Windows Resource counter and indicator description