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.

How can I get this graph:

(from Speeding up StackExchange)

It shows what percentage of web requests took longer than N seconds. I have IIS logs and LogParser installed but not sure how to proceed.

share|improve this question
add comment

2 Answers 2

This is how you can do it with LogParser and Excel:

Step 1 Create the following query and save it as "Time taken graph.sql":

SELECT QUANTIZE(time-taken, 100) AS t, COUNT(*) as count
INTO 'Time-taken-graph.csv'
FROM u_ex*.log GROUP BY t

Step 2 Run the query and export results to CSV file:

LogParser.exe file:"Time taken graph.sql"

Step 3 Open CSV file in Excel. I will use Excel 2010 as example.
Let's say your data sits in A1:B401 range:

Initial data

Put "Time" in D1 cell. Put "Percent" in E1 cell. Fill time in D column with series starting from 0 to 5 with step 0.1:

time series

Step 4 Put the following formula into E2 cell (you will need to replace 401 with your value):

=SUMIF($A$2:$A$401,">="&D2*1000,$B$2:$B$401)/SUM($B$2:$B$401)

formula

Copy the formula to all cells in E column that have corresponding time value. Set style to Percent by pressing Ctrl+Shift+%

Step 5 Finally, build line graph based on the data in D and C columns: Graph

share|improve this answer
    
When doing anything like this I use Perl to read the data (most commonly stored in a database) to create the charts in Excel. A bit of searching should turn up a number of articles on how to go about it. –  John Gardeniers Apr 22 '10 at 8:32
add comment

I wrote a python program to generate that graph using the logs generated by our load balancer and flot to draw the actual graph.

I went through a couple of iterations before I decided on that graph:

  • I started with a scatter plot (response time versus time of day) which is informative in it's own right, good for getting a good feel for the shape and variance of your traffic even if it's not particularly good for communication.

  • Then I tried a histogram, which wasn't particularly useful because of the high variance.

  • Finally I ended up with this which is based on a histogram, but is cumulative and inverted.

I'd post the code but it is so specific to what I was doing that it isn't going to help anybody. So here's an approximation of the core function:

def point(times, cutoff):
    """
        times: sorted list of response times
        0 <= cutoff < 1
    """

    size = int(len(times) * cutoff)
    return (times[cutoff], 1 - cutoff)

You then plot the (x, y) coordinates as cutoff ranges over [0,1[ using your favourite plotting library.

share|improve this answer
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.