–Summary: Lots of traffic; no server problems; Reddit loves Macs

Yesterday, Tyler’s post about the Grand Canyon Skywalk had the good fortune of hitting both the top of Reddit and the front page of Digg, one after the other.  That presented a golden opportunity to see what the traffic from each site was like.  I knew that Digg was considerably larger, but I didn’t appreciate how significant the difference was until I saw the traffic first-hand. Here’s a timeline (all times in CDT):

• April 4: Tyler releases his Grand Canyon post to the world.
• April 9 — 9:50 a.m.: First inbound referral from Joel on Software
• April 9 — 10:20 a.m.: First inbound referral from Digg
• April 9 — 10:25 a.m.: First inbound referral from Reddit
• April 9 — Mid-afternoon: Story hits the front of Reddit
• April 9 — 5:30 p.m.: Story hits the front page of Digg

Tyler’s web site is hosted on my server, so I was able to watch behind the scenes as the traffic grew.

On most normal weekdays, the server is idling along and using about 600 kbit/s of outbound bandwidth. The vast majority of that bandwidth is consumed by Bonnevilleclub. When Tyler’s story hit Reddit, the outbound traffic spiked to 1.7 Mbit/s and then relaxed a bit to about 1.0 Mbit/s. While the story was on top of Reddit, a link would come in every so often from Digg, but the Digg traffic was in the noise. That changed when the story became “popular” on Digg.

Within a few seconds of hitting the Digg front page, traffic for the blog post jumped from about a request per second to over ten per second. Outbound traffic spiked too, from about 900 kbit/s to 4.7 Mbit/s. Here’s a chart showing the traffic at the server’s network interface:

As you can see on the chart, traffic stayed extremely heavy for about 20 minutes before falling off.

The server handled the task without breaking a sweat. The only real issue was that the database kept bouncing off a concurrent connection limit, but I increased the (artificially low) limit and the problem went away.

So, what numbers can we extract from the traffic? As of noon today (the day after), there have been 76000 visits to the article from Digg, 23600 from Reddit, and 6200 from Joel on Software. Of the three sites, it seems that Reddit has the largest proportion of Macintosh fans, with 17.2% of the visitors from Reddit using Macs (Digg: 16.2%; JoS: 8.4%). Similarly, Reddit users are the most likely to be using Linux/UNIX, with 8.6% of visitors from Reddit running X11 (Digg: 4.7%; JoS: 7.5%). Rounding out the stats, 70.5% of visitors from Digg used Firefox (Reddit: 67.6%; JoS: 65.2%). These numbers are quite different from the web at large. If we assume that visitors to these sites are bellwethers for the general computing industry, Apple is poised to recapture an enormous amount of market share.

Traffic on the server is still slightly heavier than normal, with the current upstream usage at about 1.3 Mbit/s. I predict that it will return to pre-Digg/Reddit levels by tomorrow. The challenge is now in Tyler’s court: convert the one-time visitors into regulars.