The two big disadvantages are: 1) COST!!! Clustering is usually not a very cheap solution both in terms of hardware costs and software costs. For example a standard copy of WebLogic costs 10K per CPU. A clusterable copy costs 20K per CPU. This effectively doubles your licensing costs for going with a clustered WebLogic solution. Most other vendors are the same. 2) Decreased Serviceability. A cluster is harder to manage than a single server. This obviously makes sense. The more nodes you throw into the cluster, the more work it is to administer the cluster. Many clustering implementations are making good headway into this area with Auto-Discovery and Cluster-wide Deployment capabilities. [ September 18, 2002: Message edited by: Chris Mathews ]
Overhead! Like multi-tasking, clustering adds a certain extra cost to each work unit. Which is balanced by the fact that you can perform work concurrently instead of serially. The clustering equivalent to balancing the cost of overhead is the ability to schedule more efficiently between processors and the fact that loss of a processor will have minimal impact on the work.
An IDE is no substitute for an Intelligent Developer.