You are currently offline, waiting for your internet to reconnect

Optimizing Windows NT for Performance

This article was previously published under Q146005
SUMMARY
When starting troubleshooting performance problems or when trying tooptimize Windows NT, while working with Windows NT Performance Monitor itisn't always easy to choose from the large number of performance monitorobjects and counters.

This article helps you pick the right counters and objects.

Below you'll find performance checkpoints for the following possiblebottlenecks:
  • Suspected bottleneck: memory
  • Suspected bottleneck: processor
  • Suspected bottleneck: physical disk
  • Suspected bottleneck: network
  • Suspected bottleneck: network components
MORE INFORMATION

Suspected Bottleneck: Memory

Check:

Object: Memory
Counter: Pages /sec

Object: Logical Disk (location of the PAGEFILE.SYS)
Counter: Avg. Disk sec/Transfer

If the product of these two counters (equals percentage of the disk accesstime used by paging) > 10 % on a sustained basis, the system needs morememory.

Check:

Object: Memory
Counter: Pages /sec

If counter value is consistently > 5 , suspect memory.

Check:

Object: Server
Counter: Pool Nonpaged Failures

The number of times allocations from nonpaged pool have failed. Indicatesthat the computer's physical memory is too small.

Check:

Object: Server
Counter: Pool Paged Failures

Pool Paged Failures indicate that either physical memory or a paging fileis near capacity.

Check:

Object: Server
Counter: Pool Nonpaged Peak

The maximum number of bytes of nonpaged pool the server has had in use atany one point. Indicates how much physical memory the computer should have.

Suspected Bottleneck: Processor

Check:

Object: Processor
Counter: %Processor Time

If this value is consistently high (> 80%) and disk and network countervalues are low, suspect the processor.

Object: System
Counter: %Processor Time (for multi processor systems)

If this value is consistently high (> 80%) and disk and network countervalues are low, suspect the processor.

Object: System
Counter: Processor Queue Length

A sustained processor queue length > 2 , generally indicates a processorbottleneck.

Object: Processor
Counter: Interrupts/sec

A dramatic increase in this counter value without a corresponding increasein system activity indicates a hardware problem.

Object: Processes (_Total)
Counter: %Processor Time

If more than a couple of processes are contending for the majority ofthe processor time, then a faster processor or an additional processorshould be considered.

Suspected Bottleneck: Physical Disk

Check:

Object: Physical Disk
Counter: %Disk Time

If this value is consistently high and disk queue length is greater than 2,suspect the disk.

Object: Physical Disk
Counter: Average Disk sec/Transfer

A high value (values greater than 0.3 seconds) may mean that the diskcontroller is continually retrying the disk because of failures.

Object: Physical Disk
Counter: Disk Queue Length
Counter: Average Disk sec/Transfer

The Average Queue Time is the average amount of time for a disktransfer (either reads or writes) to complete. Use the following formula tofind the average disk queue time:
Avg. Queue Time = Disk Queue Length x Avg. Disk sec/Transfer

This information is a relative performance measurement and should becompared with other hard disk drivers in your system. Compute the figuresfor all logical disks in your system. The number of disk commands waitingin the queue is normally the factor that slows disk performance byincreasing the average disk queue time.

Object: Physical Disk
Counter: Disk Bytes/sec

A Disk Bytes/sec count lower than 20K may indicate that an application isaccessing a disk inefficiently.

Suspected Bottleneck: Network Components

Check:

Object: Redirector
Counter: Current commands

If this number is greater than one per network adapter, the redirectormay be a bottleneck in the system for one or more of the following reasons:

- the server with which the redirector is communicating is slower than the redirector.
- the network may be experiencing capacity problems.
- the redirector is busier than the adapter can keep up with.

If network capacity problems are identified, it may be necessary to subnetthe network in an attempt to partition network traffic.

Check:

Object: Redirector
Counter: Network Errors/sec

If any network errors are logged, check the Event Log for more details.

Check:

Object: Redirector
Counter: Reads Denieds/sec
Counter: Writes Denieds/sec

These values indicate if the remote servers are having problems with memoryallocation.

Check:

Object: Server
Counter: Work Item Shortage

An increase in Work Item Shortage should cause a change in the registryvalue(s) InitialWorkItems and/or MaximumWorkItems (depending on when theoutage occurred).

Check:

Object: Server
Counter: Raw Reads Rejected/sec
Counter: Raw Writes Rejected/sec

Rejections indicates the exhaustion of RAW work items used when busy doinglarge file transfers. The increase of the registry value RawWorkItems canpossibly solve this bottleneck.

Suspected Bottleneck: Network

Check:

Object: Server
Counter: Bytes Total/sec

If the sum of Bytes Total/sec for all servers is roughly equal to themaximum transfer rates of your network, you may need to segment thenetwork.

prodnt 3.50 3.51
Properties

Article ID: 146005 - Last Review: 02/21/2007 00:26:34 - Revision: 2.2

  • Microsoft Windows 2000 Server
  • Microsoft Windows 2000 Advanced Server
  • Microsoft Windows 2000 Professional Edition
  • Microsoft Windows NT Workstation 3.5
  • Microsoft Windows NT Workstation 3.51
  • Microsoft Windows NT Workstation 4.0 Developer Edition
  • Microsoft Windows NT Server 3.51
  • Microsoft Windows NT Server 4.0 Standard Edition
  • KB146005
Feedback