A local disk-based file system, LDFS, is an attractive way to speed up distributed applications. Local file access is much faster than accessing data on remote file servers through the network. LDFS is also scalable, as it does not rely on centralized file servers, and it exploits already existing resources (local disks) to provide storage. However, since individual workstations are less reliable and less available than file servers, LDFS must be made fault tolerant. We present an approach that integrates the LDFS with the distributed application. This is particularly suitable for mobile agent systems, because they can easily migrate to access remote files. LDFS avoids logging of individual file accesses, which are regenerated automatically from application messages. Our experiments show that the overhead of checkpointing with LDFS is generally smaller that with NFS, while access time to files decreases dramatically.

Coordination protocols are an essential part of every distributed system. In general, centralized protocols are simpler and more efficient than distributed ones. However, as a distributed system gets large, the bottleneck of the central coordinator renders protocols relying on centralized coordination inefficient. To solve this problem, hierarchical coordination can be used. This solves the scalability problem of the algorithms relying on centralized coordinations, since the performance of hierarchical coordination degrades logarithmically with the number of participating processes.

In this paper we present a mechanism that automatically organizes processes in a hierarchy and maintains the hierarchy in the presence of node failures, and addition and removal of processes in the system. The proposed scheme is simple and general, and can concurrently support multiple logical structures, such as a ring, a hypercube, or a mesh.

This paper presents a mechanism for migration of parts of computations in systems based on distributed objects or autonomous agents. In such systems load balancing and process migration can be achieved by migrating objects, or logical nodes, between participating computing nodes.

The presented migration mechanism uses minimal knowledge of the system topology and requires only a moved object to stop computing while in migration, which makes it usable in large distributed computations.

This paper presents a new checkpointing algorithm for systems using reliable communication channels. The new algorithm requires O(n + m) communication messages, where n is the number of participating processes, and m is the number of late messages. The algorithm is non-blocking, requires minimal message logging, and has minimal stable storage requirements. This algorithm is also scalable, simple, transparent to the user, and facilitates fast recovery. By introducing sutable delay in the checkpointing process, the parameter m can be made small. We also describe a variant of the algorithm that requires only O(n) messages, at a cost of O(n) additional storage for each process.

This paper also presents an efficient coordination mechanism, called the Process Order. The Process Order mechanism can be used for grouping processes in arbitrary structures in order to solve various problems, including scalability, failure detection, and coordinator election. The Process Order merchanism groups the processes transparent to the user, and automatically adjusts to the changes in system topology.

Centralized coordination protocols are simpler and more efficient than distributed ones. However, as a distributed system gets large, the bottleneck of the central coordinator renders protocols relying on centralized coordination inefficient. To solve this problem, hierarchical coordination can be used, where performance degrades logarithmically with the number of participating processes.

In this paper we present a mechanism that automatically organizes processes in a hierarchy and maintains the hierarchy in the presence of node failures, and incremental addition and removal of processes in the system. The new topology resulting from a change is computed by each process locally, without having to broadcast the entire topology to all processes. The proposed scheme can concurrently support multiple logical structures, such as a ring, a hypercube, a mesh, or a tree. It supports total order of broadcasts and does not rely on any specific system features or special hardware.

This paper proposes a new approach to rollback-recovery for mobile-agent systems, and describes its implementation in the MESSENGERS mobile agents system. The used checkpointing method allows to implement space and time efficient, user-transparent rollback-recovery in heterogenous distributed environments. Together with an efficient non-blocking system snapshot algorithm this checkpointing method is an attractive choice for implementing a roll-back recovery mechanism in the mobile agent system, because it exploits features specific to mobile agent systems during the recovery.

This paper also presents an optimization technique, called concurrent checkpointing, that increases the effectiveness of the proposed rollback-recovery mechanism.