Track: High Performance, Scalable

Abstract: The enterprise software industry is becoming increasingly dependent on the ability to scale, driven by the rapid increase in the number of cores per CPU and by the mainstream adoption of virtualization solutions such as VMWare and Xen. However, many enterprise applications and frameworks are not really built with scalability in mind. In this talk, Patrick explores the major problems surrounding scalable and concurrent architectures, including: "data coupling," in the form of shared memory and the synchronized keyword; "service coupling," in which tight implementation coupling between ostensibly separate parts of an application impedes scalability; and "performance coupling," in which performance-related concerns drive decisions to collocate business logic. We will look at how to identify likely trouble spots in existing code and discuss some techniques to eliminate or reduce them. Patrick then discusses how to design new systems to be scalable from the beginning. This section of the talk includes a survey of interesting non-Java solutions for the JVM, such as Erlang and Scala.