Architecture, Design Patterns And Load Balance In Data Based Application Partitioning

High performance, high availability, and high scalability become the three main categories to determine the system quality as the enterprise applications progress. Moreover, because people rely more and more on the fast computing ability and immediate responses, enterprise applications enter a new era of distributed environment from the standalone box episode. The classic software architecture, symmetric software architecture and asymmetric software architecture cannot the meet the current requirements. They are complex, huge, utilizing the computing resources in a bad way, and they cannot provide enough supports for high performance, high availability and high scalability.In this thesis, based on the symmetric and asymmetric software architecture, we present a new software architecture — data based application partitioning in distributed environment. It adopts the advantages from both symmetric and asymmetric software architectures, process request data according to the data attributes, and largely avoid the computing resource contention, such as database and message middlewares.Secondly, we present design patterns in data based application partitioning environment, which include Broadcast, Dispatch, Queue of Queue, XA Transactions and their comparison in performance and scalability. These design patterns are summarized from three large financial software and they provide practical, proved, robust solutions to the specific problem set.Thirdly, based on the load balance theory, we create load index and redistribution algorithm in data based partition environment. The load index in measured by database access workload - database writes and database reads. And the redistribution algorithm is well proved by our simulation program.Then, we apply data based software architecture to several large financial systems. Three case studies, including their architectures, performance and scalability test results show the benefits of data partition software architecture.Last, a conclusion of software architecture, design patterns and load balance in data based application partitioning is given, which is followed by the reference, publication list and thanks.