Mark Day
Barbara Liskov
Andrew C. Myers
Abstract
Thor is a distributed object-oriented database where objects are stored persistently at highly available servers called object repositories, or ORs. In a large Thor system, performance tuning and system reconfiguration dictate that objects must be able to migrate among ORs. The paper describes two schemes for object references that support object migration, one using location-independent names and the other, location-dependent names. The paper analyzes the performance of the two schemes and concludes that location-dependent names are the right choice for systems like Thor, where we want fast access to objects that have migrated.
- CAREY, M. J., DEWITT, D. J., AND NAUGHTON, J.F. 1993. The 007 benchmark. In Proceedzngs of the 1993 ACM SIGMOD (Washington, D.C., May), 12-21. Google Scholar
- CARSON, S., AND SETIA, S. 1992. Optimal write batch size in tog-structured file systems. In Proceedings of the 1992 Usenix File Systems Workshop, pp. 79-91.Google Scholar
- FOWLER, R. J. 1985. Decentralized object finding using forwarding addresses. Tech. Rep. 85-12-1, Dept. of Computer Science, Univ. of Washington, Dec.Google Scholar
- GRAY, J., AND REUTER, A. 1993. Transactton Processing: Concepts and Techmques. Morgan Kaufmann, San Mateo, Calif. Google Scholar
- HORNICK, M. F., AND ZDONIK, S.B. 1987. A shared, segmented memory system for an object oriented database. ACM Trans. Inf. Syst. 5, 1 (Jan.), 70-95. Google Scholar
- HwANG, D. J.-H. 1988. Constructing a highly-available location service for a distributed environment. Tech. Rep. MIT/LCS/TR410, MIT Lab. for Computer Science. Jan. Google Scholar
- JUL, E., LEVY, H., HUTCHINSON, N., AND BLACK, A. 1988. Fine-grained mobility in the Emerald system. ACM Trans. Comput. Syst. 6, i (Feb.), 109-133. Google Scholar
- KAEHLER, T., AND KRASNER, G. 1990. LOOM--Large object-oriented memory for Smalltalk-80 systems. In Readings in Object-Oriented Database Systems, S. B. Zdonik and D. Maier, Eds. Morgan Kaufmann, San Mateo, Calif., 298-307. Google Scholar
- LADIN, R., LISKOV, B., SHRIRA, L., AND GHEMAWAT, S. 1990. Lazy replication: Exploiting the semantics of distributed services. Tech. Rep. MIT/LCS/TR-484, MIT Lab. for Computer Science, July. Google Scholar
- LINDSAY, B. 1981. Object naming and catalog management for a distributed database manager. In Proceedings of the 2nd International Conference on Distmbuted Computing Systems (Paris), 31-40.Google Scholar
- LISKOV, B., DAY, M., AND SHRIRA, L. 1994. Distributed object management in Thor. In Distributed Object Management, M. T. (~zsu, U. Dayal, and P. Valduriez, Eds. Morgan Kaufmann, San Mateo, Calif.Google Scholar
- MAHESHWARI, U. 1993. Distributed garbage collection in a client-server, transactional, persistent object system. Tech. Rep. MIT/LCS/TR-574, MIT. Google Scholar
- Moss, J. E.B. 1990. Design of the Mneme persistent object store. ACM Trans. Inf. Sys. 8, 2 (Apr.), 103-139. Google Scholar
- STAMOS, J.W. 1982. A large object-oriented virtual memory: Grouping strategies, measurements, and performance. Tech. Rep. SCG-82-2, Xerox PARC, May.Google Scholar
Index Terms
- References to remote mobile objects in Thor
Recommendations
Implementing Location Independent Invocation
The problems of finding objects in large and wide-area networks where objects may change their location in volatile memory as well as on stable storage are presented. The authors discuss possible solutions and describe those adopted in the Hermes system ...
Deputy Mechanisms for Object-Oriented Databases
ICDE '95: Proceedings of the Eleventh International Conference on Data EngineeringConcepts of deputy objects and deputy classes for object-oriented databases (OODBs) are introduced. They can be used for unified realization of object views, roles and migration. The previous researches on these concepts were carried out separately, ...
Migrating Autonomous Objects in a WAN Environment
Along with the fast progress the interconnection of computer systems makes the need for enterprise-wide distributed solutions grows. These systems have to support hundreds or even thousands of sites located all over the world. The distances between the ...
Reviews
Jason Gait
Name server and remote link strategies for name resolution in the distributed object database Thor are evaluated. When a name service is used to resolve names, communication overhead may be accrued when the object is accessed. On the other hand, an efficient remote link strategy accrues communication overhead to update the link whenever an object is moved (or at least on first access after a move). The authors argue that Thor objects are accessed a lot more than they are moved, so it is more efficient to pay the cost at move time than at access time. Since the authors do not provide any measurements from Thor, I am not convinced. In a real-world distributed system, a relatively small number of objects should account for the bulk of the references. I would therefore expect that a name server strategy combined with client name caches would be an effective high-performance alternative. The advantage of a centralized (albeit replicated) name repository in a distributed database, especially from the system administration perspective, should be a factor in selecting a name resolution strategy.
Access critical reviews of Computing literature here
Become a reviewer for Computing Reviews.
Comments