This paper examines the performance characteristics of both asynchronous and synchronous parallel particle swarm optimisation algorithms in heterogeneous, fault-prone environments. Algorithm convergence is measured as a function of both iterations completed and time elapsed, allowing the two particle update mechanisms to be comprehensively evaluated and compared in such an environment. Asynchronous particle updates are shown to negatively impact the convergence speed in regards to iterations completed, however the increased parallel efficiency of the asynchronous model appears to counter this performance reduction, ensuring the asynchronous update mechanism performs comparably to the synchronous mechanism in fault-free environments. When faults are introduced, the synchronous update method is shown to suffer significant performance drops, suggesting that at least partly asynchronous algorithms should be used in real-world environments where faults can regularly occur.

Abstract — In this paper, we study parallelization of multiobjective optimization algorithms on a set of hetergeneous resources based on the Master-Slave model. Master-Slave model is known to be the simplest parallelization paradigm where a master processor sends the function evaluations to several slave processors. The critical issue when using the standard methods on heterogeneous resources is that in every iteration of the optimization, the master processor has to wait for all of the computing resources (including the slow ones) to deliver the evaluations. In this paper, we study a new algorithm, where all of the available computing resources are efficiently utilized to perform the multi-objective optimization task independent from the speed (fast or slow) of the computing processors. For this we propose a hybrid method using Multi-objective Particle Swarm optimization and Binary search methods. The new algorithm has been tested on a scenario contaning heterogeneous resources and the results show that not only the new algorithm performs well for parallel resources, but also when comparing to a normal serial run on one computer.

This paper describes a distributed particle swarm optimisation algorithm (PSO) based on peer-to-peer computer networks. A number of modifications are made to the more traditional synchronous PSO algorithm to allow for fully decentralised, scalable and fault-tolerent operation. The modified algorithm uses staggered propagation of objective-space knowledge between sub-swarms to eliminate the need for a centralised data store. Analytical test functions are used to examine the performance of the proposed algorithm and its variations in comparison with a basic synchronous PSO implementation. The results clearly show the feasibility of decentralised particle swarm optimisation.

Abstract — This paper proposes a method for designing EMC shielding enclosures using a peer-to-peer based distributed optimisation system based on a modified particle swarm optimisation (PSO) algorithm. This optimisation system is used to efficiently obtain optimal solutions to a shielding enclosure design problem by sampling only a small fraction of the total problem space. Such a system would find use in environments where large scale computing resources are not available, such as smaller engineering companies, where it would allow computer-aided design by optimisation using existing resources with little to no financial outlay. the y-axis, with a perforation radius of between 2 and 35 mm. Constrained to feasible solutions, this gave approximately 1100 possible configurations. The P2P-PSO system was evaluated using 8 nodes with a sub-swarm size and iteration count of 2 and 10 and 1 and 20, with both configurations requiring only 160 simulations be performed. Using these configurations, the P2P-PSO system was able to obtain a reasonable approximation of the real optimal solution set (also known as the Pareto front) for this problem as shown in fig. 1, greatly reducing the required computational time. I.

Abstract—Population-based, multi-objective optimisation algorithms are increasingly making use of distributed, parallel computing environments. In these cases it is a commonsense precaution to consider the possibility of a variety of failures. In particular, errors caused by response failures are more prone to arise than in homogeneous parallel computers. While masking errors using redundant computation is simple and reasonably reliable, it is expensive in terms of the computing resources required. An alternative approach is presented that uses a Byzantine agreement methodology, utlising only results already computed. In computational experiments it has a demonstrated ability to correct errors, and salvage useable results from unreliable, distributed computing environments. With increasing reliance on computing resources provided and operated by external agencies, error detection and correction can be expected to become more important to a range of applications. Index Terms – Fault tolerance, distributed computing, multiobjective optimisation.