09-WebLogic性能调优概要.ppt

* While it is obvious that sockets knock RMI’s lights out, don’t go away just yet. It would be just as easy to concoct a test that showed almost negligible differences based on the amount of computation the server does. Sockets are the primitive building blocks for distributed computing. As such they are more efficient than higher layer protocols like RMI. The performance gap between RMI and Sockets is much more pronounced if the server workload is light. As the compute load increases, the network I/O bottleneck imposed by upper layers decreases.. In the same way that data size impacts sockets performance, RMI performance is strongly influenced by the complexity of serializing the remote method’s parameters and return value. Complex objects are more expensive to serialize and, will weigh down RMI speed. * Threading allows programmers to multitask within their applications. However, there are often sections of code that should not be executed by more than one thread at a time. These sections of code are known as critical sections. Critical sections contain code that changes common variables, updates a table, or writes to a file, etc. The execution of the critical sections of code must be mutually exclusive. Synchronization allows programmers to make sure that only one thread executes critical sections at a time. A critical section could be a block of code, a method, or a class. Synchronization makes sure that only one thread executes a critical section at a time and all other threads block and wait. Synchronization causes a bottleneck since only one thread can execute at a time. Synchronization is key to Java programming but it should not be overused. When making architecture decisions between two classes select the simplest one. Some JDK classes use synchronization and if you don’t need that support it would just add overhead to your application. Many Java classes are used together to develop complex objects. Such a complex object may handle synchronization calls