O-Raid: experiences and experiments

Abstract

We present the O-Raid distributed object-oriented database system, which is built by extending an existing relational database system called RAID. We compare the O-Raid data model with those of other systems and describe its design and implementation. We present the experimental data of studies performed to measure the overheads involved in supporting objects. Experiments included queries on objects in a single site, as well as queries on replicated composite objects spanning multiple sites. We also discuss the issues related to replication, indexing and fragmentation of objects. We conclude with our experiences and lessons learned in developing the O-Raid system.<>