10.9. Summary

This chapter on making Flex and BlazeDS applications scalable is a collection of varied recommendations and viewpoints. The chapter started with the important topic of clustering and explained how multiple logical BlazeDS instances aggregated logically can help service a larger number of Flex clients concurrently.

The chapter then delved into efficient tricks for managing data. Gzip-based data compression and binary object data formats were recommended over uncompressed textual data.

The discussion then moved to the topic of Java NIO-based asynchronous nonblocking sockets. Java NIO, with its selector-based intermediary, makes socket connections scalable without the need for extensive multi-threaded program manipulation.

After robust connection and nonblocking IO, the tenets of service orientation that apply to RIA were elaborated and recommendations were discussed at a conceptual level.

Caching and resource pooling are important methods of increasing scalability and availability. These methods exist in every area of system development from operating systems to application software. The section on caching briefly demonstrated how signed RSLs can be effective. The section on resource pooling explained database connection and JMS resource pooling.

The final section ended with a quick discussion of workload distribution, which reiterated that creating thicker clients should be preferred over creating thin clients as in traditional web applications.