2 of 14

"State of the Union"
It's probably safe to say that most CF developers start out using session-scoped variables to manage client state. Session variables are fast and easy to use. In addition, there seems to have always been some confusion over the differences between client and session variables. Since session variables are most often discussed in the newsgroups, and are most frequently found in the various code examples in help screens, newsgroups and books, the implementation of session variables can be considered the most common method of client state management for new to intermediate CF developers.

Furthermore, it's also safe to say that "newbie" CF developers have no inherent need to work with any other method of client state management, since they're working primarily within a single-server environment. Since a single-server development or production platform poses no problems for developers using session variables for client state management, the session scope is an excellent choice.

A Clustering "Primer"
There are times, however, when using session variables can get you into trouble - specifically when running the application in a clustered-server environment. Before I address this problem, let's talk about what a clustered server environment really is. While an in-depth discussion of clustering is beyond the scope of this article, I will offer this brief "primer."

Configuring a cluster involves installing either hardware or software into the network infrastructure that will load-balance a Web site or Web application across multiple servers. This provides scalability: as usage of a Web site or Web application grows and server traffic increases, more horsepower can easily be added to the mix to compensate.

This article considers a specific and common configuration: ColdFusion 4+ with BrightTiger's ClusterCATS. In the CF/ClusterCATS configuration, the ClusterCATS software component connects two or more ColdFusion Web servers together, allowing them to share the work of processing and serving up Web pages. It also provides an element of fault tolerance; should one of the machines become unavailable (because of a hardware, software or network failure), the "load" is shifted to the remaining machine(s). All of this clustering functionality is (or should be!) transparent to the users.