Performance Analysis For Distributed Real-Time Software With Aspect-Oriented Technology

Performance, which is an important characteristic of a software product, is critical to the success of today's software systems, especially to the complex distributed real-time software. Performance problems are most often due to inappropriate architectural choices rather than inefficient coding. By the time the architecture is fixed, it may be too late to achieve adequate performance by "fix-it-later" approach. Software Performance Engineering (SPE) advises that performance must be designed into software from the beginning.Performance, which is one of the non-functional concerns, crosscuts the whole distributed real-time systems, and cannot be well encapsulated into a single module with object-oriented technique. Aspect-oriented technique is able to separate and modularize crosscutting concerns, such as performance, during the whole software development process. Therefore it is good for performance modeling and measurement.Whereas, by the time performance becomes the new focus of attention in aspect-oriented software development, the performance issues resulted from aspect have to draw our attention. Because aspect-oriented technique provides higher modularization, which will lowered efficiency of systems, especially in distributed real-time systems.Aspect-oriented Performance Engineering Framework (AOPEF) takes pros and cons depicted above into consideration. AOPEF, mainly including performance modeling and measurement, considers that aspect-oriented technique has impact on performance, as well as we can make use of it to construct a plug-and-play performance probe with higher efficiency, accuracy, and least impact on software architecture and behavior.In performance modeling phase, AOPEF provides Performance Scenario Charts (PSC) for performance parameters specification. PSCs not only describe the executionpaths of critical scenarios, but also specify the layered performance values and resource requirements. Then PSCs are transformed into Aspect-oriented Layered Queuing Networks (AO-LQN) Model, which merges several concurrent scenarios into one model to analyze the performance of distributed systems under multi-workload circumstances, and which has capability of identifying bottlenecks and priority inversion with Use Case Maps (UCM) appending, and which is able to output various performance evaluated values with LQN solver. In addition, an overhead matrix can provide us more accurate values.In performance measurement phase, AOPEF advises to embed a performance probe for measurement in the process of aspect-oriented software design and performance modeling. That is why aspect-oriented technique is superior to object-oriented technique in performance analysis.The last but not least, a distributed web application paradigm is made to demonstrate the details of performance modeling and measurement in AOPEF.