The SHRIMP project investigates how to construct high-performance servers with a network of commodity PCs and commodity operating systems. The cost of such a multi-computer server is substantially less than a commercial, custom-designed multi-computer. The goal is to study how to build such a system to deliver performance competitive with or better than the commercial multi-computer servers.

The research project consists of several components: user-level, protected communication, efficient message-passing, shared virtual memory, distributed file system, performance measurement, scalable 3D graphics, and applications.

Princeton University Computer Science Department is building a parallel computer using PCs running Linux as the processing elements. The first was a simple two-processor prototype that used a dual-ported RAM on a custom EISA card interface. A recent prototype will scale to larger configurations, using a custom interface card to connect to a "hub" that is essentially the same mesh routing network used in the Intel Paragon (see http://www.ssd.intel.com/paragon.html). Considerable effort has gone into developing low overhead "virtual memory mapped communication" hardware and support software.

Collaborators
University of Tromsø, Norway; Princeton University, USA

Contact
Kai Li
Princeton University, USA
li@cs.princeton.edu

Otto Anshus
University of Tromsø, Norway
otto@cs.uit.no

http://www.CS.Princeton.EDU/shrimp/