A comparative study of structured and un-structured remote data access in distributed computing systems

Abstract

Recently, the use of distributed computing systems has been growing rapidly due to the result of cheap and advanced microelectronic technology. In addition to the decrease in hardware costs, the tremendous development in machine to machine communication interfaces, especially in local area networking, also favours the use of distributed systems. Distributed systems often require remote access to data stored at different sites. Generally, two models of access to remote data storage exist: the un structured and structured models. In the former, data is simply stored as row of bytes, whereas in the latter, data is stored along with the associated access codes. The objective of this thesis is to compare these two models and hence determines the tradeoffs of each model. First of all, an extended review of the field of distributed data access is provided which addressing key issues such as the basic design principles of distributed computing systems, the notions of abstract data types, data inheritance, data type system and data persistence. Secondly, a distributed system is implemented using the persistent programming language PS-algol and the high level language C in conjunction with the remote procedure call facilities available in Unix(^1) 4.2 BSD operating system. This distributed system makes extensive use of Unix's software tools and hence it is called DCSUNIX for Distributed Computing System on UNIX. Thirdly, two specific applications which employ the implemented system will be given so that a comparison can be made between the two remote data access models mentioned above. Finally, the implemented system is compared with the criteria established earlier in the thesis. keywords: abstract data types, class, database management, data persistence, information hiding, inheritance, object oriented programming, programming languages, remote procedure calls, transparency, and type checking.