Unify: A Scalable, Loosely-Coupled, Distributed Shared Memory Multicomputer

The objective of the Unify project is to develop a scalable distributed shared-memory (DSM) multicomputer linking hundreds or thousands of high-performance machines in geographically distant locations. There are five main areas of research being addressed by the Unify project:

New Scalable DSM Abstractions:

We are developing new DSM abstraction that relax consistency and synchronization requirements to produce better concurrency and mask the high latencies of a wide-area system.

Reliable Multicast Communication:

We are developing a scalable reliable multicasting protocol called TMTP to reduce bandwidth requirements and ensure timely reliable delivery of data.

Performance Monitoring Tool:

We are developing a graphical user interface that monitors distributed applications as they execute and helps users fine-tune performance.

Adaptive Object Support:

We are developing a reactive object library that will allow applications to automatically adapt themselves to the current network environment to attain better performance.

Checkpointing/Fault Tolerance:

We are adding support for Checkpointing and recovery in a system with relaxed consistency requirements.