Windows performance Monitoring

See also:http://www.cnblogs.com/updoogis/archive/2010/11/19/1881970.html

Cpu

Processor:% Processor Time

The current processor time for all processes [i.e. CPU utilization], if the value continues to exceed 95%,cpu is the bottleneck. % Processor time value =100%-idle process ratio [that is: idle thread]],CPU It's okay to execute a free thread.

System Processor Queue Length

Memory Performance Counters

Memory:available Mbytes

The amount of memory that can be used. If the indicator's data is small, there may be a memory problem with the system

Memory:pages/sec

Indicates the number of pages fetched from disk due to hardware page faults, or the number of pages written to disk to free working set space due to a page fault. Generally if the pages/sec continues above hundreds of, then you should study the paging activity further. There may be a need to increase memory to reduce the need for page breaks. A large value of pages/sec does not necessarily indicate a memory problem, but may be caused by a program running a memory-mapped file. A high ratio of counters indicates excessive paging.

Memory:page faults/sec and Memory:page reads/sec counters measure memory performance.

Page faults occur when the virtual memory page referenced by the process is not in memory, and Memory:page Faults/sec represents the number of page faults. If a page is already in main memory, or if it is being used by another process that is sharing this page, then the page will not be paged in from disk.

Memory:page Reads/sec is a read disk that extracts the number of pages required to resolve page faults. In general, these two counters reflect the frequency of access to the disk, and the lower the value, the less the response time.

Network Performance counters

Main View Network Interface object

Network interface:bytes received/sec

How many bytes data received per second, combined with bytes total/sec analysis

Network interface:bytes sent/sec

How many bytes data per second are sent, combined with bytes total/sec analysis

Network interface:bytes total/sec

The total number of machines accepted and sent bytes recommended not to exceed 50% of the bandwidth

Network interface:packets/sec

The number of packets per second, according to the actual data volume size, no recommended threshold, the data combined with bytes total/sec analysis

Current Bandwidth number of bits per second

Network Utilization NIC utilization = ((total bytes\sec * 8)/current bandwidth) * 100

Disk Physical Disk\Disk Transfers/sec IOPS

About disk response time (disk Response) entry: Ms MS

1.Avg. Disk Sec/transfer: Shows the average time of each IO processed by the storage side.

2.Avg. Disk Sec/read: Shows the average time of each read IO processed by the storage side.

3.Avg. Disk sec/write: Shows the average time of each write IO processed by the storage side.

These entries are the first to be viewed in the performance analysis, and Disk Response time directly determines the level of service the storage system has for the application. Typically the user feels the performance problem and is also because disk Response time on the disks is up. The following is a database application as an example, give some reference disk Response time, if the disk response in the corresponding range is considered acceptable, otherwise you need to see why. Of course, depending on the production environment and the application of the situation also need to be treated differently.

Data files:

1-5MS, more ideal.

? 6-20ms, acceptable.

? More than 20ms, there will be performance issues that require a solution.

Database transaction log:

1-2ms, more ideal.

? 2-6ms, acceptable.

? 6-15ms, need to be optimized.

? More than 15ms, there will be performance issues that require a solution.

Data Warehouse:

Data Warehouse due to the application of the IO type characteristics, usually a single IO will be relatively large, so that the response time of the IO will increase, usually for the data warehouse and reporting system, 25-30MS will be more healthy value.

About IOPS Entries:

1.Disk transfers/sec: Shows the total iops on disk, visually showing the throughput of IO per second on the corresponding LUN.

2.Disk reads/sec: Shows the number of read IO per second on the disk.

3.Disk writes/sec: Shows the number of write IO per second on the disk.

Depending on the reference value processed by the IOPS, we get the reference IOPS for the different physical disks. However, the situation in the actual production environment is more complex, due to the existence of the storage array cache mechanism, the IOPS calculation for a specific LUN involves a cache hit, a healthy storage environment, the IOPS is much higher than the sum of the physical disk theoretical values. But it's not that the higher the IOPS, the better the storage performance problem, the first thing to note is the disk Response time mentioned above. If the IOPS is at a low point and does not reach the theoretical value of the physical disk, but disk Response time is high, there is a problem with the storage side.

The following is a list of the reference IOPS values for several different physical disk types:

? 15000rpm HDD 120–180 IOPs

? 10000rpm HDD 100–150 IOPs

? 7200rpm HDD 80-100 IOPs

? Enterprise SSD Disk 2000–5000 IOPs

About throughput (MB/s) entries:

1.Disk bytes/sec: Displays the total amount of data transferred on the disk, in bytes.

2.Disk Read bytes/sec: Shows the amount of read data transferred on disk.

3.Disk Write Bytes/sec: Shows the amount of write data transferred on disk.

The display of this value is converted to MB and divided by 1024^2. The point to mention here is that the IOPS and bandwidth display are not fully proportional. Typically, for small IO cases, the focus is on IOPS, which is throughput (MBs) for large IO scenarios.