Designing Distributed Sites
By John N. Stewart
Having spent the past four years helping organizations design distributed Web sites, I've seen various forces in the Web industry converge. Sites that were once fully within an organization's control now are composed of contractual relationships with other organizations, like advertisement vendors, stock-quote providers, and commerce clearing houses. Served from multiple locations or multiple organizations, distributed sites are quickly becoming the status quo.
The operational and strategic concerns of an organization's Web site change when it becomes distributed. The site's backend functionality becomes more difficult to manage, and in some cases, is accessible only via a business relationship. And with organizations beginning to depend on their Web sites as mission-critical applications, site plans now require disaster-recovery and load-balancing features to deliver maximum uptime and performance. Distributed Web sites require proper design, security, technology, and procedures. In short, knowing the risks and pitfalls is essential to limiting them.
During the design phase, it's important never to lose sight of the individual machines and components that make up the site. It's too easy to believe that a site is working because some HTML code from the home page was successfully retrieved by the monitoring service. In truth, your Web site is only working if the service it offers is flawless. If the home page references images at another site, or loads a page from another server, those services had best be fully operational.
The challenges presented by distributed sites can be placed in five categories: content management, content distribution, distributed monitoring, site development, and the adaptation of existing security policies and procedures.<>
