Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

As per this answer, I've configured perfmon to show

  • Memory / Pages Input/sec
  • CPU / CPU Time (%)
  • Physical Disk / Average Queue length

(Names might be slightly different on an English version of Windows). Now I see these average values:

  • Memory: 74.613 (1.000)
  • CPU: 16.642 (1.000)
  • Disk: 0.160 (100.000)

How do I interpret these values? CPU is simple (16.6% usage).

But how about disk? Is that 16 requests every second? Or 0.16? Or 0.0016? That doesn't seem right; the LED is flashing madly.

And page faults: Is that 74 page loads/sec?

For the fun of it, I've added "Physical Disk / Bytes read/sec" and "Physical Disk / Bytes written/sec". Here I get 235478.228 and 30568.626 respectively with a factor of 0.0001. Does that translate to 235MB/s read (implausible with a desktop harddisk) or 235 Bytes/s? Again the LED on the case indicates it must be much more.

Thanks a lot for clearing this up.

[EDIT] One thing which I figured out: The "factor" is to scale the value to be able to display it in the graph. The values below the graph (current, average, min, max) are absolute (or unscaled).

[EDIT2] Sorry, I mixed up the factors for memory and queue length.

[EDIT3] I'm on Windows XP/SP3.

And for those people who have been looking for the "Explain" button: 1. Click on "Add" (new indicator). In the dialog, there is an "Explain" button which tells you something about the currently selected indicator.

And a message to MicroSoft: If you supply a list box to select one option out of a whole lot, make that widget a bit bigger, okay? Scrolling wastes valuable human CPU power.

share|improve this question
    
What OS are you on? –  Jim B Jul 21 '09 at 16:39
add comment

3 Answers

up vote 3 down vote accepted

There is a free tool called PAL (Performance Analysis of Logs) http://www.codeplex.com/PAL which is intended to help read and understand performance logs. The author is an engineer at Microsoft. The tool has thresholds for many of the values recorded by perfmon. It might help.

share|improve this answer
    
The tools looks nice but my version of perfmon doesn't allow to save data to files, so I can't use PAL :( –  Aaron Digulla Jul 21 '09 at 14:40
    
What OS are you on? Perfmon in every NT-derived OS that I'm aware of lets you save counter data to files. –  Evan Anderson Jul 21 '09 at 23:34
    
Okay, I found that further down in the tree on the left, I could create a protocol ("Leistungsindikatorenprotokolle"). My problem with the graphical view of the performance is that there is an item "Save data as..." in the context menu but it's always disabled. –  Aaron Digulla Jul 23 '09 at 12:00
add comment

The "Pages Input / sec" counter indicates that your machine is pulling in 74.613 memory pages per second from the pagefile (the (100) indicates that the value is being scaled in the display by 100). You need to graph "Page Reads / sec" counter and see if it is following the "Pages Input / sec" counter closely.

If you are seeing excessive paging, add the "Page Faults / sec" counter from the "Process" object for each process instance listed (besides _Total) and move thru them looking for the process(es) that are paging excessively. Bear in mind that "Page Faults / sec" is showing both soft and hard page faults for each process combined into a single total.

That disk queue number means that .16 requests were waiting in queue for the disk to become available per second. That number should normally be lower than the number of physical spindles in a RAID array, or under 2 for a single drive. This would be in line with you having a reasonably fast disk but having some excessive paging occurring.

Your disk reads are in line with the amount of paging you're seeing. Get to the bottom of the paging and life will be good.

The "Explain" button in Perfmon is your friend. Take a look at him.

share|improve this answer
    
I have 4GB (3.5 can be used), 1.5GB free. What's a sane value? Or how much "off" is .74? And where would I find the explain button? I checked the tooltip of every button but the only thing that remotely matches the description is the online help (which doesn't). –  Aaron Digulla Jul 21 '09 at 13:48
    
When you go to add counters, there is an Explain button, highlight the type and counter you're interested in. –  SqlACID Jul 21 '09 at 13:54
    
Sorry, I mixed up the factors for memory and queue length. See my edits. –  Aaron Digulla Jul 21 '09 at 13:56
    
Evan- pages input being hi all by itself means you have gobs of FREE memory. You have to compare it to page reads to see if all the inputs are for requests by default windows is going to try to read back in as many pages as possible- from the description "When a page is faulted, the system tries to read multiple contiguous pages into memory to maximize the benefit of the read operation. Compare the value of Memory\\Pages Input/sec to the value of Memory\\Page Reads/sec to determine the average number of pages read into memory during each read operation." –  Jim B Jul 22 '09 at 1:27
    
@Jim: You're quite right. Apparently today is my day to be a complete idiot. I was thinking about "Page Reads / sec" when I said what I did, but talking about "Pages Input / sec". –  Evan Anderson Jul 22 '09 at 1:34
add comment

Evan is right, the explain button should help, but more importantly make sure you are using the right tool to get the answer you want. You haven't said what you are looking at perfmon for. If you are simply wondering why the disk is spinning use process monitor under XP or the disk section of reliabilty and performance under vista/windows 7. If you are having a memory problem pages/sec would let you know as it's the sum of pages in and out to disk (remember that generally windows swaps FROM Disk not TO disk during a memory shortage and only pages out userdata- not applications)

share|improve this answer
    
I know exactly what's wrong: the virus scanner is hogging the disk. But I need something to show my boss so we can do something about it. You know, something that will make a non-tech-type go ballistic just by looking at it. The curves already look good but I'd like to avoid some embarrassing stammering when I have to answer the odd question. –  Aaron Digulla Jul 21 '09 at 13:58
    
The buttons in my window read (in order): New, Clear, Show Current something, Show Report Data, Graph, Histogram, Report, Add, Delete, Mark, Copy, Paste, Properties, Stop, Snapshot, Help –  Aaron Digulla Jul 21 '09 at 14:05
    
Are you just trying to get rid of the antivirus on your system or stop a scheduled scan? Is there a performance issue, or is this just a question about why the light is on? I've not used an AV software that doesn't pause on user activity (except when a scheduled scan starts- thanks symantec!). Disk activity in and of itself isn't an indicagtor of poor system performance, but I'm sure you could make charts showing activity. If you are on vista there is actually disk I/O priority which makes activity almost meaningless when looking for a bottleneck if the i/o priority is set to low –  Jim B Jul 21 '09 at 16:39
    
Currently, my computer freezes every now and then, Explorer popup menus take up to 30 seconds to appear and when I type in this comment box, I see 2-3 characters per second appear. The whole system is on the brink of unusable, yet the CPU indicator is a mere 10% so my boss is going to say "What are you talking about? Looks good to me!" –  Aaron Digulla Jul 22 '09 at 8:12
    
After a day of investigation, I think it's the schedules scan but I need more than a feeling in the meeting with the security guys. –  Aaron Digulla Jul 22 '09 at 8:13
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.