Transparency and reflection in distributed systems

ABSTRACT

The recursive composition of systems to form functionally equivalent transparently distributed systems is an important paradigm for constructing distributed systems. The extent to which such recursive transparency can be achieved depends crucially on the semantics and functionality offered by the underlying systems. It is therefore important that systems should be designed so that their functionality scales gracefully in a distributed environment.In order to build a transparent extension to a system, it is necessary to be able to intercept its basic operations and extend their meaning to a distributed environment. This requires the underlying system to have a clean structure with well-defined interfaces and the reflective capability necessary to intercept and extend operations crossing those interfaces. Thus, both reflection and transparency are important aspects of the design of extensible distributed systems.It is possible to make transparent extensions to object-oriented systems built on top of micro-kernel architectures but the lack of reflective capabilities in the current generation of object-oriented programming languages can make this unnecessarily awkward. More research is required to develop languages whose computational model and implementation are a better match for the underlying platforms which support them.

REFERENCES

• 1: J.P. Black, L.F. Marshall, B. Randell, "The architecture of Unix United", Proc. IEEE, Vol 75, No 5, May 1987, pp. 709-718
• 2: B. Randell, "Recursively structured distributed computer systems", in Proc IEEE Symposium on Reliability in Distributed Software and Database Systems, 1983, pp. 3-11
• 3: R. Sandberg et al, "Design and implementation of the Sun network file system", in Proc. USENIX conference, June 1985, pp.119-130
• 4: James H. Morris , Mahadev Satyanarayanan , Michael H. Conner , John H. Howard , David S. Rosenthal , F. Donelson Smith, Andrew: a distributed personal computing environment, Communications of the ACM, v.29 n.3, p.184-201, March 1986 [doi>10.1145/5666.5671]
• 5: Marc Shapiro, Soul: An Object-Oriented OS Framework for Object Support, Proceedings of the International Workshop on Operating Systems of the 90s and Beyond, p.251-255, July 08-12, 1991
• 6: Robert J. Stroud, Naming issues in the design of transparently distributed operating systems, 1987
• 7: Pattie Maes, Concepts and experiments in computational reflection, ACM SIGPLAN Notices, v.22 n.12, p.147-155, Dec. 1987
• 8: Fridolin Hofmann , Peter Schlenk , Thomas Eirich, Encapsulation and Interaction in Future Operating Systems, Proceedings of the International Workshop on Operating Systems of the 90s and Beyond, p.241-246, July 08-12, 1991
• 9: M.Shapiro, P. Gautron and L. Mosseri, "Persistence and migration for C++ objects", Proceedings of the 3rd European Conference on Object-Oriented Programming, 1989, pp. 191-204
• 10: David L. Detlefs , Maurice P. Herlihy , Jeannette M. Wing, Inheritance of Synchronization and Recovery Properties in Avalon/C++, Computer, v.21 n.12, p.57-69, December 1988 [doi>10.1109/2.16189]
• 11: Santosh K. Shrivastava , Graeme N. Dixon , Graham D. Parrington, An Overview of the Arjuna Distributed Programming System, IEEE Software, v.8 n.1, p.66-73, January 1991 [doi>10.1109/52.62934]