Abstract Scalable computer systems, including clusters and multi‐cluster grids, require routine exchange of information about the state of system‐wide resources among their nodes. Gossip‐based algorithms are popular for providing such information services due to their simplicity, fault tolerance and low communication overhead. This paper presents a randomized gossip algorithm for maintaining a distributed bulletin board among the nodes of a scalable computer system. In this algorithm each node routinely disseminates its most recently acquired information while maintaining a snapshot of the other nodes' states. The paper provides analytical approximations for the expected average age, the age distribution and the expected maximal age for the acquired information at each node. We confirm our results by measurements of the performance of the algorithm on a multi‐cluster campus grid with 256 nodes and by simulations of configurations with up to 2048 nodes. The paper then presents practical enhancements of the algorithm, which makes it more suitable for a real system. Such enhancements include using fixed‐size messages, reducing the number of messages sent to inactive nodes and supporting urgent information. The enhanced algorithm guarantees the age properties of the information at each node in the configurations with an arbitrary number of inactive nodes. It is being used in our campus grid for resource discovery, for dynamic assignment of processes to the best available nodes, for load‐balancing and for on‐line monitoring. Copyright © 2009 John Wiley & Sons, Ltd.

Proportional (fair) share schedulers are designed to provide applications with predefined portions of system resources. Single node operating systems use context-switch (preemption) to dynamically allocate the CPU(s) to running processes. This paper presents an online algorithm for proportional share allocations of nodes in a cluster, in a fashion that resembles a single-node system. The algorithm relies on preemptive process migrations for dynamic allocations of nodes to users. The paper presents the algorithm and its performance on a MOSIX organizational Grid with 60 nodes. We show that proportional share allocations can be achieved in a relatively short time (minutes).

One of the main methods for achieving fault tolerance in distributed systems is recovery of the state of failed components. Though generic recovery methods like checkpointing and message logging exist, in many cases the recovery has to be application specific. In this paper we propose a general model for a node state reconstruction after crash failures. In our model the reconstruction operation is defined only by the requirements it fulfills, without referring to the specific application dependent way it is performed. The model provides a framework for formal treatment of algorithm-specific and system-specific recovery procedures. It is used to specify node state reconstruction procedures for several widely used distributed algorithms and systems, as well as to prove their correctness.