Performance Monitor

in performance monitor, there many counters choose.

for windows server, think there 3 important categories monitor: disk, processor, memory.

what important counters each of these 3 categories system administrator needs monitor?

thanks.

question
commonly used performance monitor (perfmon) counter thresholds?

answer

this article lists common perfmon counters descriptions , thresholds. threshold values listed here meant use general 'rule-of-thumb' , each should interpreted in context of specific performance issue @ hand.

when using these values, keep following in mind:

1. the threshold values provided below averages, not min/max , useful when looked @ within meaningful time-period following points clarify.
2. what time range captured in perfmon data set?
3. extreme highs , lows within time-range of capture can result in less-useful averages. in these cases, try narrow time range , @ counter values around time(s) when poor performance observed.
4. in general, considered genuine bottleneck, given counter's threshold must exceeded either on average, or on frequent basis, during time period being analyzed.
5. this not meant comprehensive nor definitive reference.

object	counter	description	normal range	threshold
.net clr memory	% time in gc.<instancename>	% time in gc percentage of elapsed time spent in performing garbage collection (gc) since last gc cycle. counter indicator of work done garbage collector on behalf of application collect , compact memory. counter updated @ end of every gc , counter value reflects last observed value; not average.	<30%	>=30%
memory	free system page table entries	the number of page table entries not in used system.  counter displays last observed value only; not average.	>8,000	<8,000
	pages/sec	the rate @ pages read or written disk resolve hard page faults. counter primary indicator of kinds of faults cause system-wide delays...	0-15	>20
physical disk	avg. disk queue length	the average number of both read , write requests queued selected disk during sample interval.	<=2*totalnumberofdisks	>2*totalnumberofdisks
	avg. disk sec/read	the average time, in seconds, of read of data disk.	4-8ms	>15ms
	avg. disk sec/write	the average time, in seconds, of write of data disk.	<=10ms	>50ms
process	%processor time.<instancename>	% processor time percentage of elapsed time of process threads used processor execution instructions.	<=50*totalnumberofcores	>50*totalnumberofcores
	privatebytes.w3wp.exe	the current size, in bytes, of memory process has allocated cannot shared other processes.	without /3gb switch: <=800mb with /3gb switch: <=1,800mb	without /3gb switch: >800mb /3gb switch: >1,800mb
processor	% processor time._total	percentage of elapsed time processor spends execute non-idle thread.	<80	>80
sqlserver: buffermanager	buffer cache hit ratio	percentage of pages found in buffer cache without having read disk ... because reading cache less expensive reading disk, want ratio high...	>98	<90
	free pages	total number of pages on free lists.	>=640	<640
	page life expectancy	number of seconds page stay in buffer pool without references.	>=300	<300

 

refer to: http://www.symantec.com/business/support/index?page=content&id=howto9722

hope article helps.

thanks

zhang

Windows Server  >  Windows Server General Forum