Quantifying the Slashdot Effect
October 24, 2008 4:54 PM
Subscribe
I'm trying to figure out if a server I'm working on can withstand being linked to by very high traffic websites. Can anyone tell me the basic traffic characteristics of, say, a front page link from: Drudge Report, Yahoo! Buzz, Huffington Post, Slashdot, Reddit, or Digg? I'm using ab (ApacheBench) and I'd like to have some idea of how many requests, and at what level of concurrency, I should be trying to make this server capable of pushing out. Thanks!
I already know what to do (log and observe, cache cache cache, tune and tweak db and web server configuration, etc) and am experienced in that process. I just need some typical numbers to help me figure out when I'm done!
posted by evariste to computers & internet (5 comments total)
2 users marked this as a favorite
really well, which means ten thousand requests for a page, issued 500 at a time. Some of the page requests are slow, but 75% take under 2 seconds and 99% are under 17 seconds. The slowest one took 45 seconds, and one request couldn't be satisfied. Throughout this, the server's load average remained well below 1. A good start? Just average? I have no basis on which to judge this.It would be cool if someone could tell me, e.g., "Slashdot's traffic comes in spurts of about 5000, maybe 20 concurrently, until you've seen about two hundred thousand pageloads" because that way I could write a script that simulated a Slashdotting (or whatever).
posted by evariste at 5:01 PM on October 24, 2008