регистрация / вход

MONITORING AND OPTIMIZATION Essay Research Paper PERFORMANCE

MONITORING AND OPTIMIZATION Essay, Research Paper PERFORMANCE MONITOR AND NETWORK MONITOR PERFORMANCE MONITOR Windows NT Object Counters in Performance Monitor

MONITORING AND OPTIMIZATION Essay, Research Paper

PERFORMANCE MONITOR AND NETWORK MONITOR

PERFORMANCE MONITOR

Windows NT Object Counters in Performance Monitor

OBJECTPURPOSE

CacheL2 cache performance

Logical diskMass storage performance, including network storage

MemoryMemory performance and usage

ObjectsProcess and tread counts

Paging fileVirtual memory usage

Physical diskHD performance

Process Performance of executing processes

Processor Processor(s) performance

SystemWindows NT performance

Thread Individual thread performance

You will also see objects for each network service installed.

MONITORING FOR PROCESSOR BOTTLENECKS

MAJOR PROCESSOR RELATED COUNTERS TO WATCH:

Processor: %Processor Time – How busy the microprocessor is. Processor is a bottleneck if sustained >80%

Processor: Interrupts/sec – Rate of service requests from peripheral devices. If you have a high rate of interrupts/sec with no corresponding hardware functions then you may have a bad piece of hardware. Should be between 100 and 1000. Spikes to 2000 are ok.

System: Processor queue length – Number of threads waiting to be serviced. >2 then processor is a bottleneck. Queue length will always be zero unless you are also monitoring a thread specific counter (Context switches/sec is number of threads switched/sec by the processor and is good to use).

MONITORING THE DISK FOR BOTTLENECKS

IF THE DISK LIGHT IS ON ALL THE TIME, YOU NEED MORE RAM

More Ram will increase the size of the disk cache and decrease page swapping to disk and will increase the apparent speed of the disk.

Logical disk vs. Physical disk

Two different objects in Performance monitor.

Logical disk measures performance of stripe sets, volume sets and mapped network drives.

Physical disk measures real transfers to and from actual hard disk or RAID set. Used to compare disks and to provide specific information about a disk.

MAJOR DISK RELATED COUNTERS TO WATCH:

Memory: Pages/sec shows the number of memory pages swapped out to disk per second. Useful with %disk time: modify the size of the page file and watch the result of these two disk counters.

%Disk time does not directly measure the disk. It shows how much time the processor is spending servicing disk requests. Use with Processor: %Processor Time to determine if the disk is eating up the processor.

Disk bytes per second shows how fast your disk is. Copy a big file and see if your disk is way fast or a piece of junk.

Average disk bytes per transfer shows how big the average transfer is. Larger transfers are more efficient.

Current Disk queue length shows how much data is waiting to be transferred to disk. A long queue means you need a faster disk.

NOTE: You must enable disk counters, they are disabled by default and will cause a degradation of about 2% in performance due to processor load of the counter.

To enable disk performance counters: type diskperf -y at the command prompt to enable auto counter startup at boot. Reboot.

To disable disk performance counters: type diskperf -n at the command prompt to disable auto counter startup at boot. Reboot.

NETWORK MONITOR

Is a trimmed down version of what ships with SMS.

Is installed by adding Network monitor tools and agent through the services tab of the network icon in control panel.

Can provide real time and cumulative saved data.

FOUR MAIN SECTIONS

Bar graph in real time.

Session statistics shows cumulative data about conversations taking place on the network in real time.

Station statistics shows information on each conversation. You must specify the machine (station) you want to monitor and is cumulative for the monitoring period.

Summary statistics are cumulative and show network, captured, per second, NIC (MAC), and NIC error statistics.

Data can be filtered by protocol, computer address, or protocol properties. Filter by computer address to identify a NIC sending data frames when not in use (bad NIC).

Netmon can be dual password protected and will let you identify other instances of Netmon running on the network. This is to prevent unauthorized users from capturing data to which they do not have permission to access.

??? Conflict ???

Sybex Network Press: MCSE NT Server 4 Study Guide says, ?The limited version of network monitor that ships with Windows NT Server does not support promiscuous mode. Therefore, it can capture only packets sent to the server or to all stations.?

But?

Windows NT Online Help for Performance Monitor says, The Nework Monitor agent collects statistics from the computer?s network adapter card by putting it in promiscuous mode.?

ОТКРЫТЬ САМ ДОКУМЕНТ В НОВОМ ОКНЕ

ДОБАВИТЬ КОММЕНТАРИЙ [можно без регистрации]

Ваше имя:

Комментарий