Hi I am learning about Windows performance troubleshooting via Perfmon, and was hoping for some help. One of the most common tickets we recv from other IT departments is that their server is running slowly. Generally, what I do is: 1. Launch Task Manager and check CPU usage 2. Within Task Manager, look at available Physical Memory However, I realise this is just the tip of the iceberg. For instance, there can be network issues, disk issues etc, which I guess only Perfmon will show. So I have been reading these articles here: http://www.windowsnetworking.com/articles_tutorials/Windows-Server-2003-Performance-Tuning.html# http://adminfoo.net/2007/04/windows-perfmon-top-ten-counters.html But I am still having trouble with the following points - 1. For various counters, Perfmon will give me the Max, Min and Average. Is there anyway to check these figures against MS recommended values? Otherwise, these are just numbers that don't mean anything! 2. Each counter has a scale. Ok so a scale of 1.00 is easy to understand, but - for example - Avg. Disk Queue Lght has a scale of 100.00. What does this mean? Do I have to multiple the values for Max, Min, Avg etc by 100 or divide by 100? 3. I understand Pages/Sec should be less than 20. Am I right in thinking that paging is effectively using the hard disk instead of RAM, and paging > 20 means more RAM should be added? 4. What is the best counter to measure delays in the disk subsystem? 5. Let's say I want to measure performance over a 12 hour period, and then be able to go back to that "report", find out a spike and tie that into a specific time, is this possible? If so, how? 5. What is the best counter (if Pages/sec isn't) to measure memory usage? Any help appreciated guys!

1. For various counters, Perfmon will give me the Max, Min and Average. Is there anyway to check these figures against MS recommended values? Otherwise, these are just numbers that don't mean anything! Generally speaking you need to create a baseline of the server to compare against. While there are a simple general guidelines (disk queue should not exceed 2 per disk, etc), most of these are very superficial. Also, most performance issues will be with individual service (IIS, SQL, File Server, etc), the performance counters for these are very application dependent. People spend years (you will too) learning how to dig in to discover the tell-tell signs of a performance issue of services. Keep working at it. Disk and Memory are pretty universal, if you can troubleshot those, then you have 50% of all issues covered. 2. Each counter has a scale. Ok so a scale of 1.00 is easy to understand, but - for example - Avg. Disk Queue Lght has a scale of 100.00. What does this mean? Do I have to multiple the values for Max, Min, Avg etc by 100 or divide by 100? The scale is just how it displays on the graph. It will make more sense to see it happen, but here is a general explanation. Since the AVG (or current) disk queue should not exceed 2. If your disk queue was 2 for 20mins and you have a scale of 1, then 2 would be shown on the graph...not much of a bump, not noticable at all...you might miss it. But if have a scale of 100 then 200 would display on the graph, big BUMP, VERY noticable....that is what you want...given the importance of the counter. 3. I understand Pages/Sec should be less than 20. Am I right in thinking that paging is effectively using the hard disk instead of RAM, and paging > 20 means more RAM should be added? This counter is the most miss understood counter, and Microsoft should remove it from the default display. Pages/Sec very rarely represents any meaningfull information, mostly because of the erradic behaviour most admins think it is indicating some performance issue for various reasons, when in fact it is just showing you NORMAL behavior. I ignore this counter most of the time....or at least until you better understand what it is meant to represent. 4. What is the best counter to measure delays in the disk subsystem? Avg disk response time is the best indicator....but acceptable performance varies with speed and type of disk and amount of CACHE. Avg Disk sec/Read Avg Disk sec/Write In general i like this, but the Mid may be a little high....these number vary alot depending on various factors (CACHE, RAID, etc) and alot depending on load. High Performance 15k Read and Write <.010 sec Mid Performance 10k Read and Write <.020 sec http://networkadminkb.com/Shared%20Documents/Disk%20IO%20Tutorial%20Part%201.aspx 5. Let's say I want to measure performance over a 12 hour period, and then be able to go back to that "report", find out a spike and tie that into a specific time, is this possible? If so, how? Log to File or SQL Server, then use a the graph to determine time. 5. What is the best counter (if Pages/sec isn't) to measure memory usage? Available Memory in MB/GB...tells you how much you have left. More Informaton: http://networkadminkb.com/kb/Knowledge%20Base/Windows2003/How%20to%20performance%20tune%20a%20Windows%202003%20Server.aspx

Hi Thanks for the great reply, much appreciated!! Just some follow up if that is ok.. >The scale is just how it displays on the graph. It will make more sense to see it happen, but >here is a general explanation. Since the AVG (or current) disk queue should not exceed 2. If >your disk queue was 2 for 20mins and you have a scale of 1, then 2 would be shown on the >graph...not much of a bump, not noticable at all...you might miss it. But if have a scale of 100 >then 200 would display on the graph, big BUMP, VERY noticable....that is what you want...given >the importance of the counter. Ah ok, so the values displayed in Last, Average, Min, Max, Average etc are all correct, the scale is just for graph purposes? Regarding Pages/Sec, MS recommend that this value should be less than 20. Would you agree, or advise just to ignore this counter anyway until I understand OS and hardware more? :) >In general i like this, but the Mid may be a little high....these number vary alot depending on >various factors (CACHE, RAID, etc) and alot depending on load. >High Performance 15k >Read and Write <.010 sec >Mid Performance 10k >Read and Write <.020 sec Sorry, not sure what the values refer to there? 15k refers to??... >Log to File or SQL Server, then use a the graph to determine time. I can't see how to "Log to File", would you know where I need to go to specify the time period etc? I've right clicked on the System Monitor, but can't see this option? Sorry to be a pain, I've tried to google this as well, but can't see anything! Thanks again so much for your help!

>>Ah ok, so the values displayed in Last, Average, Min, Max, Average etc are all correct, the scale is just for graph purposes? Correct >>>Regarding Pages/Sec, MS recommend that this value should be less than 20. This is what MS recommends (in my own words): 0-20 is ideal, but its only a concern if >80. Source below. http://technet.microsoft.com/en-us/library/cc782186(WS.10).aspx 0–20. (Unhealthy if greater than 80; probably indicates not enough RAM.) Why they (MS) even have 0-20 is beyond me. Why not just say >80 is of concern?? Get why i think this is misunderstood? I have yet to trace any performance issue back to memory using this counter as my justification, others counter will indicate the real problem, this counter MAY lead you to an actual memory performance issue (if you really understand it), but for most users not likely. >> Would you agree, or advise just to ignore this counter anyway until I understand OS and hardware more? :) Ignore for now, work on understanding memory architecture and the different types of memory. >>Sorry, not sure what the values refer to there? 15k refers to??... 15k (15,000), 10k (10,000) are speeds of hard drives in RPM. .001 is 1 milisecond > Fastest .009 is 9 miliseconds .010 is 1 hunderth of a second (aka centisecond) .020 is 2 hunderth of a second > Slowest http://en.wikipedia.org/wiki/Millisecond >>>I can't see how to "Log to File", Assuming XP/2003 Start Perfmon (System Monitor), On the Right click the + next to Performance Logs and Alerts Select Counter Logs, on the RIGHT under the default log (System Overview), right click and Select New Log Settings... Type Name, and then configure as needed. Use Help...Help Topics if you have any more questions about Perfmon.