Web Caching

Chapter 9. Cache Clusters

A cache cluster is a group of separate caching proxies configured to act like a single server. In other words, even though there are many machines in the cluster, clients and users perceive it as a single unit.

A cluster is different from a hierarchy or mesh in some subtle ways. Primarily, the members of a cluster are typically located together, both physically and topologically, for example, with servers in the same room and on the same subnet. Since a cluster appears as a single system, a single organization has administrative control over all the machines. A hierarchy, on the other hand, often involves multiple locations and multiple organizations.

Why would you need a cluster? Many companies use cache clusters to serve or cache more pages and provide redundant service. Let’s say you’re already using a certain caching product you really like, but your traffic load is increasing, and the service is becoming slow. You need to upgrade, but you don’t want to discard your current product. You can double your capacity if you buy a second server and create a small cluster. As mentioned earlier, if a cache is down, users may not be able to surf the Web. In a cache cluster, if one server fails unexpectedly, the other(s) can absorb the additional load.

This chapter covers three different reasons for clustering. We’ll start with ways to provide redundancy and not worry about additional capacity. Then we’ll focus on techniques for distributing load among multiple ...