A Distributed Rule Engine Based On Message-Passing Model To Deal With Big Data

As a good way of knowledge representation and inference in Artificial Intelligence, rule engines have been extensively studied and applied both in academia and industry. Nowadays the developing of Big Data not only has offered a wider range of application but also has raised a great challenge since the low computational efficiency of rule engines when dealing with a large amount of data. What is worse, when the size of a problem continues to grow, the efficiency becomes much lower quickly. Although there are various of research trying to address this issue, almost all of them have different kinds of limitation, such as inadequate optimization, lack of universality and so on.On the basis of former research about rules engines with knowledge of parallel and distributed computing, in this dissertation we propose a new approach to implement a distributed rule engine based on message-passing model. The approach aims to make use of multiple machines in a computing cluster to deal well with a large amount of data in a parallel and distributed way.In this dissertation, firstly we describe details of the design, which involve meth-ods of mapping the core of a rule engine on message-passing model, of distributing the system and of allocating knowledge with other kinds of data. Secondly, after introduc-ing several problems of performance found in a verifying prototype, we discuss special solutions. Thirdly, we describe details of implementation, which are mainly about the architecture of the whole system as well as main functions and APIs of divided mod-ules. Finally some experiments that we have conducted are introduced and discussed to verity the high performance of the implemented rule engine when dealing with big data, which benefits from parallel and distributed computing based on message-passing model.