Speed and Reliability, Part Two
By Mitch Wyle
In part one of this article on load balancing for high-demand Web sites (see "Business Developer," Web Techniques, January 2000), we restricted our discussion to "large egress" bandwidth requirements for multimedia -- that is, Web sites that need to broadcast or download large files. In this second installment, we'll examine the more common problem of high transaction volume. All Web-site owners want to have "the high traffic problem." We'd all like to own and administer 50-million-hit-per-day sites that make lots of money, or effectively use the Web to promote the purpose of our enterprise. Unfortunately, sites that get a sudden flood of traffic when an advertising campaign launches can melt down under the sudden load and even cause businesses to fail as a result of their own successful advertising!
Of course, all Web teams want to provide a positive experience for the end user. Among the ingredients for such an experience are a reliable, low-latency, end-to-end network connection that leads to fast loading times. Ideally the speed of operation should approach that of point-and-click local desktop applications. And, especially for "dot com" businesses (as opposed to "dot org" clubs with no budget for high-end hardware) the goal is to maintain this wonderful end-user experience during peak traffic times when dozens or hundreds of simultaneous hits to the same URLs occur.
Some of the speed problems are caused by local ISP congestion at the end user's side of the Net.
