Plurix is a native cluster Operating System (OS) written in Java. The cluster nodes communicate via Distributed Shared Memory (DSM) and the consistency of the DSM is guaranteed by restartable transactions and optimistic concurrency control. Standard Internet protocols lead to a smooth integration of Plurix nodes into existing network environments and open up the world wide web to Plurix. In this paper we present the design and implementation of network protocols for the management of Plurix transactions and DSM on top of IP. We discuss strategies to support long lived TCP connections in an environment where transactions need to be short to minimize collision probability. Finally we present preliminary performance results of an early prototype which was shown at the CeBIT 2000 trade fair.
[1]
David Abramson,et al.
Implementing a large virtual memory in a Distributed Computing
,
1985
.
[2]
Stefan Traub,et al.
Speicherverwaltung und Kollisionsbehandlung in transaktionsbasierten verteilten Betriebssystemen
,
1996
.
[3]
Alexander Thomasian.
Distributed Optimistic Concurrency Control Methods for High-Performance Transaction Processing
,
1998,
IEEE Trans. Knowl. Data Eng..
[4]
Kai Li,et al.
IVY: A Shared Virtual Memory System for Parallel Computing
,
1988,
ICPP.
[5]
J. T. Robinson,et al.
On optimistic methods for concurrency control
,
1979,
TODS.