Open Collections

Distributed object management in Raven

University of British Columbia

Raven is an object-oriented distributed system designed to support research in building fault-tolerance distributed systems using an object model. Invocations on Raven objects are executed within a transaction mechanism with provisions for application controlled relaxations. Central to Raven underlying system support is the Object Manager which is responsible for secondary storage management of objects, concurrency control, transaction management, objects distribution and migration, recovery from failures and virtual memory management. Raven objects are encapsulated into object spaces and primary memory is simply used as a cache to object spaces in secondary storage or in remote machines. The physical layout of an object space in primary memory is identical to its layout in secondary storage. An object space represents a unit of concurrency control and recovery in Raven. Raven design integrates virtual memory management, secondary storage management and transaction management into one computation model implemented inside an Object Manager. This allows a highly efficient implementation of the Object Manager. The major work of this thesis is devoted to the design and implementations of the Object Manager in Raven.

Text

2010-08-24

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/27674

Computer Science

University of British Columbia

Graduate