An analysis of the “slashdot effect”
I’ve been in a somewhat unique position this month: I’ve been slashdotted twice, with the second time occurring within nine days of the first. Some find this to be an enviable position, and others would be annoyed by it. It seems reasonable, then, that I would feel both ways. Some brief background, though it’s probably a superfluous formality: the slashdot effect is what happens when a website gets posted to Slashdot, and the sheer mass of the incoming readers to that site typically causes that site to disappear from the Internet for a brief time because it simply cannot handle the number of visitors, especially if it is a media-heavy site. The effect is so common, there is even a Wikipedia entry on the topic. Many slashdotters pride themselves on this fact, and why shouldn’t they? There isn’t any other site on the Internet that can crash webservers just by mentioning them on the front page, that I am aware of…
The first time I was slashdotted wasn’t a real slashdotting, rather, I got the secondary traffic that was shunted my way by the first site as I was mirroring a file for them. This first time, I was quite pleased with the results, as I enjoy doing things like this for the communities that I participate in.
The effects of the slashdot effect started very slowly until the site that recieved the full force of it picked me up as a primary mirror, and then traffic exploded. I was quite astounded since I was one layer of abstraction away, and it still blew the numbers away that I put up when I first mirrored the Episode III trailers. I’ll give you some numbers now, since geeks love numbers. The Star Destroyer story was posted on July 4:
Figures are bandwidth usage/page impressions:
- July 3: 63.03MB/200
- July 4: 61.47GB/3,232
- July 5: 31.23GB/1,719
- July 6: 5.59GB/475
- July 7: 2.38GB/326
Bandwidth usage never dipped below 1GB for the remaining days. When Edda added me as Mirror 2 for the “huge” version, traffic jumped: 3,416 referrals, and that number still climbs slowly even now. As a funny aside, the night before I had put in a trouble ticket to have my server upgraded from a 10Mbps to a 100Mbps port, because one of my users was planning on distributing a some files to multiple people simultaneously, and it would be nice for him to have a fat 100Mbps pipe to do it on. I hadn’t been planning on mirroring any files for slashdot (or anyone else) for the near future. Well, I saw Edda’s video, and I just had to mirror it since I (somehow) managed to get a copy of the large version. You can read the brief exchange between the iWeb tech and myself in full here, but essentially, he happened to be a slashdot reader, and saw my site and then saw my trouble ticket from the night before, and upgraded me right away. It was a funny and cool coincidence. I highly recommend the guys at iWeb, they kick ass.
Here are the MRTG graphs from July 4:
Looking at these numbers is pretty impressive, I think. I didn’t think I’d ever see traffic like this for months/years. Little did I know what was to come in a brief nine days. The first pseudo-slashdotting was nothing in comparison.
2 Comments »
RSS feed for comments on this post.
| TrackBack URI
You can also bookmark
this on del.icio.us or check the cosmos
[...] I have written an article on the phenomenon known as the Slashdot effect, having survived it twice in the month of July alone. I’ve analyzed traffic patterns and referrals, and I’ve got graphs and numbers: all the things a good computer dork enjoys in his spare time. An excerpt for your reading enjoyment: The effects of the slashdot effect started very slowly until the site that recieved the full force of it picked me up as a primary mirror, and then traffic exploded. I was quite astounded since I was one layer of abstraction away, and it still blew the numbers away that I put up when I first mirrored the Episode III trailers. I’ll give you some numbers now, since geeks love numbers. The Star Destroyer story was posted on July 4: [...]
Pingback by polyscience.org » A meta-analysis of the “Slashdot effect” — July 15, 2005 @ 11:08 pm
[...] I’m getting rid of my dedicated server in the next two weeks and moving everything over to a $9.99 shared plan with Dreamhost. It is more than adequate for my needs, and it will save me $110 a month. While I would certainly go with a dedicated box if my websites required it, there’s just no reason for it at this point in time. (I would, however, heartily recommend iweb.ca to anyone looking for a solid, responsive, friendly, and inexpensive dedicated server with plenty of bandwidth.) [...]
Pingback by rianjs.net » Things I’m planning or may change in the future — June 15, 2006 @ 4:00 pm