Rule Mechanism In Distributed Environment

As a compact,loosely coupled,naturally parallelizable declarative language,rule language can replace intricate imperative languages in distributed applications,greatly reducing the development cost of distributed applications.Therefore,the research of rule mechanism in distributed environment is of great significance to further promote the development of distributed applications.However,the existing distributed rule system has redundant derivation in the data persistence mode,and there are redundant synchronous calls in the cooperative mode,which reduces the efficiency of rule calculation and is not easy for users to write and understand.In addition,in a distributed environment,there are often cases of network congestion or open circuit,and the existing synchronization mechanism is seriously affected by network disconnection.And in a distributed environment,it is common for multiple processes to update the data set at the same time,and the existing incremental maintenance algorithm cannot process multiple updates concurrently.In response to the above problems,this thesis has done the following research work:(1)Add explicit synchronous calls to the rules and further consider the efficiency of the synchronization mechanism in a distributed environment.A synchronization algorithm that supports dynamic adjustment of communication modes is proposed to improve the efficiency of the synchronization mechanism in distributed environment.The algorithm decouples the communication mode from the network topology by introducing a node coding method,and proposes an asynchronous adjustment operator(AAO)to adjust communication pattern at a low cost.The local continuous tree(LCT)structure is designed to reduce the synchronous communication distance of the nodes.Finally,the dynamic local continuous tree(DLCT)algorithm is implemented by combining the LCT structure and the AAO operator,and the efficiency of the synchronization mechanism is improved by the DLCT algorithm under the condition of network disconnection.(2)Use storage mode persistence and further consider the incremental maintenance efficiency of update of persistent data.An incremental maintenance algorithm that supports concurrent processing of multiple updates is proposed.The algorithm records auxiliary information by constructing a hierarchical iterative model(HIM),and designs two defined state update operations to support multi-threaded concurrent update processing.The algorithm proposes a multi-version EDB storage structure(MVEDB)to reduce the storage cost of EDB during concurrent update.Finally,the parallel incremental maintenance algorithm PIM is implemented by combining HIM and MVEDB,and the efficiency of the incremental maintenance algorithm is verified by experiments.(3)The distributed rule system is designed and implemented,and the rule compiler and stream data processing mechanism are implemented.The display synchronous calling method and the storage mode of persistent data are realized.The system designed the function module to unify the synchronization and aggregate function calls,improve the system expansion ability and reduce the user learning cost.It combines batch data and stream data processing mechanisms.The rule calculation in the incremental maintenance process is separated,and the rule-related execution plan and the algorithm-related driving process are decoupled,which reduces the complexity of the rule compiler.Finally,the efficiency of the proposed method is verified by experiments.