On July 4th 2011, my article about time appeared on Hacker News and I averaged almost 1 QPS over the next 24 hours. Here's what the webserver did:
And here's what the CPU did:
Static files, people.
On January 19th 2013, it appeared on Reddit and again on Hacker News, although with a less link-bait-like title this time. Over the next two days or so, the article got 108k hits from 63k unique clients, plus 3k hits from 400 known bots:
And here's the matching CPU graph:
39k hits came from 18.104.22.168, which resolves to boris.pinboard.in. The netblock 22.214.171.124/20 belongs to prgmr.com which is the same VPS provider I was using at the time.
The URL submitted to Reddit was /time/?v=1 which is bogus, because the article is a static HTML page which doesn't take any query parameters. Other referrers glued on even more ridiculous parameters.
Less than 2% of clients came in on IPv6. That makes me sad. We can do better, folks:
Some rough demographics:
Reddit drove most of the traffic by volume:
Like Jason Scott already told you, URL shorteners are a terrible idea.
I configure Apache to log the time taken to serve the request (%D in LogFormat) and the distribution for my article looked like this:
Request latency: min: 0.624 msec 25%: 1.789 msec 50%: 1.923 msec 75%: 2.163 msec 90%: 2.293 msec 95%: 2.344 msec
The above is with mod_deflate involved. Here's what it looks like without compression:
Request latency: min: 0.151 msec 25%: 0.521 msec 50%: 0.560 msec 75%: 0.581 msec 90%: 0.607 msec 95%: 0.627 msec
My entire website (being a handful of small static files) is served out of memory.