A Distributed Data-Storage Service for Supporting Group Collaborations TR92-044

The architecture and implementation of a distributed data-storage service is presented. Requirements for this design were deterruined from our research on computer and communications systems to support groups of users collaborating to solve complex problems such as the design of large software systems. The data model supported by the system is much richer than that of conventional file systems and is capable of directly supporting hypertext applications such as (interactive) navigational browsers. Some important issues addressed in this design include scale, performance, concurrency semantics, access protection, location independence, and replication (for fault-tolerance). We show that even with the richer functions supported, a high level of performance and scalability can be achieved. The key to this result is our application of design principles (e.g., distribution of computationally-intensive work, caching, minimal client-server interaction) found in some conventional distributed file systems. The design also avoids the necessity for more complex mechanisms such as distributed transaction processing.