Comet Scalability

Greg Wilkins has a great post on Comet benchmarking and scalability testing. He says that although a Chat application has become the defacto example for demostrating Comet’s capabilities, it may not be the most accurate example for testing Comet’s scalability characterstics. He suggests that a stock update example may be possibly more accurate. In the same post he talks about the modifications done in the Jetty /Comet implementation to effectively manage long polling and effectively handle high message rates. For example he shows in detail how an average footprint of a 157kb request footprint is reduced to 40kb in the case of a Comet supporting Servlet in the modified Jetty instance.