Check NT server performance 

---

Introduction

If you are experiencing performance problems (for example, agents are experiencing delays or polling intervals are not being met), run the Performance Monitor (located in the Administrative Tools menu) in the logging mode (as opposed to charting) with all counters selected. Set logging for every 120 seconds and save the configuration.

To obtain a sample of the server workload, run monitoring for a full working day. To study the log files, monitoring should be stopped and switched to view the chart mode. Load the data that was collected by selecting the Data From menu item from the Options menu.

The critical objects are as follows:

• Processor�% Processor Time (for each processor instance): should not stay above 70% most of the time.

• System�ProcessorQueueLength: the sustained value for this should be no larger than 2. A value higher than 2 indicates a bottleneck in the system.

• Memory�Pages/sec: if Pages/sec is greater than 10, then memory is constrained. Your system is seriously degrading if Pages/sec is greater than 20.

• Memory�PageFaults/sec and CacheFaults/sec: if CacheFaults/sec is greater that PageFaults/sec, then there is too much paging. You should increase your system's memory.

• Paging File�the Paging File should typically be set to the physical memory of the NT server plus 12MB. You should also distribute it among the physical disks in the system separate from the Windows NT system files.

• Disks�by default, disk counters are disabled. To enable disk counters, enter diskperf -y.

  You should try to attain the following values:

  • PhysDisk: %DiskTime (<85 to 90%)

  • PhysDisk: CurrentDiskQueueLength (<2)

  • PhysDisk: AvgDiskSec/Transfer (<0.3)

  • PhysDisk: AvgDiskBytes/Transfer (>20K)

• Network Interface Card�BytesTotal/sec: a 10Mbit Ethernet segment has a maximum throughput of 1.2Mbit.