Ever have a performance problem, but don’t know what performance counters to collect or how to analyze them? The PAL (Performance Analysis of Logs) tool is a new and powerful tool that reads in a performance monitor counter log (any known format) and analyzes it using complex, but known thresholds (provided). The tool generates an HTML based report which graphically charts important performance counters and throws alerts when thresholds are exceeded. The thresholds are originally based on thresholds defined by the Microsoft product teams and members of Microsoft support, but continue to be expanded by this ongoing project. This tool is not a replacement of traditional performance analysis, but it automates the analysis of performance counter logs enough to save you time. This is a VBScript and requires Microsoft LogParser (free download).
My take on the tool:
For those of us out there that don’t have to deal with performance data on a daily basis I see a few options to help troubleshoot performance issues on your servers. 1) If your using 2k3 use SPA. 2) If you’re running 2k8/Vista use data collection sets. 3) Collect analyze your own perfmon captures.
Now you might want to look into this tool. I found the tool simple to use and it’s really a four step process. The web page created for the analysis has a plethora of info and links to the codeplex site for more info. Sweet.
Really in the end it’s just a time saver. After collecting performance data on a server you need to analyze that data. This entails opening the log file, adding the counters that you’ve collected and finding out if any of the counters are above any thresholds (deemed by you). This tool does that analysis for you. It comes out-of-the-box with some predefined thresholds defined as high according to the MSFT consulting/development but those can be adjusted to whatever suits your fancy.
Once you get everything installed its time to do some analysis. It comes with some threshold templates for AD, System Overview, IIS, SQL, Exchange, etc (see pic) You point the app at the performance log you’ve captured during your perf issue, choose a threshold template to your liking, answer some basic questions, add the form and execute:
Once it completes it generates a webpage with the analysis information you desire. The webpage shows you alerts for activity that it finds suspect and graphs for the different areas of interest. I can’t paste all the pics/info in here as it is quite lengthy depending on the interval you provide. But this definitely seems like a tool that could be handy down the road. Looking at the web page it looks really similar to SPA, but with graphs provided via the Office Web Components add-in. For example here is how I could find out LDP was using too much CPU:
First I found the alert which said that something was being excessive and I clicked on the link (sorry for the blurriness):
Then I found LDP consuming the CPU:
Add it to your bag of tricks, hope it helps.