4 ms·
It is loading Javascript from her site. She probably has Apache KeepAlive on with a maximum number of connections far insufficient for the number of people cur
by patio11 16y ago
It is loading Javascript from her site. She probably has Apache KeepAlive on with a maximum number of connections far insufficient for the number of people currently trying to request files. They all get into a line and request their file, get it almost instantly, then keep the connection open for 2 ~ 3 seconds waiting for another request to come in... while several thousand people behind them have their requests time out. Javascript included in the HEAD of a document will block rendering until it is either downloaded and parsed completely or times out. Thus, it doesn't matter if Google's CDN gives you the HTML in 5 milliseconds -- your user-visible experience is dominated by her server's timeout on the Javascript.
Apache KeepAlive: it kills more blogs than cancer.
For learning about this general subject, I recommend the YSlow presentations.
- colmmacc 16y agoHi there, I'm the administrator of the box running the blog. I'm also colm@apache.org, one of the authors of Apache httpd. KeepAlive's are enabled on the host, but with a deliberate timeout of 1 second. The listen backlog is 25, which can contribute to a backlog looking like the above but it isn't. The hosted ended up getting wedged due to poor MySQL contention, there is a limit of just 80 server processes (and hence connections - we're running pre-fork on this host), and they were each getting blocked on MySQL reads. That's been sorted now, and some better caching, compression and more put in place. It's a 1Ghz, 1GB RAM box is a colo that got slammed by HN, Reddit and Techcrunch in one go - to an uncached wordpress instance. I'm surprised it did as well as it did ;-)