Visualizing performance from the server logs

by olegkikin • 16 years ago

41 4 16 years ago

jws• 16 years ago

Very nice. Interesting how the horizontal tiers in the black are stable across traffic changes (with one exception). I presume we are seeing several different types of queries combined.

nlo• 16 years ago

Good approach. Any chance of your sharing the code used to generate these graphs from the %D-formatted Apache logs?

foxtrot• 16 years ago

Nicely done, a good clean approach utilising current logs so no increase on server overheads by running the app, that I can tell.

sarabob• 16 years ago

Prompted me to run a quick version over our weblogs. You rapidly end up with big black splodges if you squish the time axis. I overcame this by multiplying the pixel colour by 0.8 for each hit to get a heatmap-style image.

Example images (first one on simple db hits; second one shows both fragment cached and non-cached page generation): http://img13.imageshack.us/img13/7290/logdots2.png and http://img375.imageshack.us/img375/4869/logdots1.png (DOS at around 6pm)

Takes around 2 mins on a pentium D (!) to generate both graphs from a single webserver's daily log, approx 1.3M log lines. I thought it would take much longer given the read/set requirement for each dot, so I'm pleasantly surprised.

Visualizing performance from the server logs

Related Stories