Roger J. Loader

Middleware enhancements for metropolitan area wireless Internet access

Abstract The effectiveness of mission critical wireless applications such as those deployed in mobile commerce depends to a large extent on network service quality. Based on the observation that wireless networks, particularly in metropolitan areas, are subject to pockets of lowered quality and disconnectivity, we propose mechanisms to detect and adapt to these ‘trouble spots’. In metropolitan area wireless networks, we postulate that the majority of network problem areas are caused by location and environmental factors. End applications that are susceptible to brief disconnectivity when entering these trouble spots can exploit pre-emptive adaptation at the protocol level to overcome potentially serious problems at the application level. Based on this premise, we propose a mechanism that gathers semantic data pertaining to specific geographic areas which cause communication problems. Prior knowledge of such locations can be used by higher-level communication protocols to pre-emptively adapt, thereby avoiding undesirable effects at the application layer. To facilitate the categorization of trouble spots, we propose a series of wireless service evaluation metrics. We report on our experiences with using these metrics and present initial results from an experimental evaluation of their effectiveness.

Skewed Crossover and the Dynamic Distributed Database Problem

The automatic self-management of large, distributed databases is a significant problem area for providers of global management information systems and services. Finding a way of dynamically balancing changing load over a number of globally distributed servers can be an arduous task, particularly when communications costs and overheads are also considered. Previous work has shown that this problem can prove a difficult search space to negotiate for Genetic Algorithms. This paper introduces a skewed form of 2-point crossover which appears to give exceedingly encouraging results, particularly on scenarios which have previously been categorised as ‘problematic’. Whilst the advantages of this form of crossover may prove to be problem (or even solution) specific, it is likely to be of use in problems where sub-sequence information is an important feature of schemata (such as scheduling problems), or where optimal solutions contain repetition of either individual or short sequences of alleles across numerous gene positions. The technique effectively provides an additional, orthogonal source of genetic diversity, apparently reducing the need for either excessive initial population size or high levels of mutation. It is also shown to be effective as part of the mutation operator in Simulated Annealing.

Investigation of a characteristic bimodal convergence-time/mutation-rate feature in evolutionary search

The use of evolutionary algorithms to determine optimum load and data distribution in large distributed databases has been investigated in earlier publications both by the authors and others. This paper reports on some interesting results arising from comprehensive examination of the performance profile of various techniques we have investigated on this problem and others. In particular, we see that when too little mutation is available to the system, the number of evaluations that the algorithm is able to exploit before premature convergence occurs seems near linearly proportional to population size, regardless of evaluation (time) limit, selection strategy, or other features. More interestingly, however, as mutation is increased, there seem to exist characteristic peaks and troughs in the tuned performance landscape indicating an optimal mutation rate independent of population size; this is a trough between the two peaks in a robust bimodal feature in the curve of convergence time against mutation-rate. These features are demonstrated over a range of evaluation limits, algorithm designs, and application landscapes. The continued re-appearance of the bimodal feature leads us to postulate that it may be a relatively problem-independent feature of evolutionary search, with general application to parameter tuning issues.

The Agreement Problem Protocol Verification Environment

The Agreement Problem Protocol Verification Environment (APPROVE) for the automated formal verification of solutions to agreement problems is presented. Agreement problems are characterized by the need for a group of processes to agree on a proposed value and are exemplified by group membership, consensus and leader election schemes. Generally it is accepted by practitioners in both academia and industry that the development of reliable and robust solutions to agreement problems is essential to the usability of group communication infrastructures. Thus, it is important that the correctness of new agreement algorithms be verified formally. In the past, the application of manual proof methods has been met with varying degrees of success, suggesting that a less error prone automated tool approach is required. Furthermore, an observation made during a review of such proofs is that a significant amount of effort is invested into repeatedly modeling re-usable themes. The APPROVE project addresses these issues by introducing a usable Spin based framework that exploits the potential for model re-use wherever possible.

Collaborative Group Membership

In this paper we present a novel approach to fault-tolerant group membership for use predominantly in collaborative computing environments. As an exemplar, we use the Collaborative Computing Transport Layer which offers reliable atomic multicast capabilities for use in collaborative environments such as the Collaborative Computing Frameworks (CCF). Specific design goals of the approach are the elimination of processing overhead due to heartbeats, support for partial failures and extensibility. These goals are satisfied in an approach, termed Collaborative Group Membership (CGM), which uses a quiescent weak failure detector and two election based algorithms to form consensus on the membership of a failing group. Failure detection operates through a reliable multicast primitive and as such eliminates the need for explicit keep-alive packets; thus in a failure free environment, CGM imposes no overhead.

Mobile Wide Area Wireless Fault-Tolerance

This paper presents work-in-progress that is developing a novel fault-tolerant mechanism for use in mobile wide area wireless networks. As a developmental platform, we are using the Ricochet service which offers ubiquitous metropolitan scale wireless network coverage in several major US cities. We postulate that the majority of network failures in infrastructures such as Ricochet are caused by environmental factors. From this, we propose a GPS based mechanism that intelligently gathers semantic data pertaining to specific geographic areas (or trouble spots) which cause communication problems. To facilitate the categorisation of trouble spots, we propose a list of suitable metrics to analyse the status of a wireless connection. Finally, we experimentally evaluate their effectiveness.

An election based approach to fault-tolerant group membership in collaborative environments

In this paper we present a novel approach to fault-tolerant group membership for use predominantly in collaborative computing environments. As an exemplar, we use the Collaborative Computing Transport Layer which offers reliable atomic multicast capabilities for use in collaborative environments such as the Collaborative Computing Frameworks (CCF). Specific design goals of the approach are the elimination of processing overhead due to heartbeats, support for partial failures and extensibility These goals are satisfied in an approach which uses an IP multicast failure detector and two election based algorithms. By basing failure detection on IP multicast, the need for explicit keep-alive packets is removed, thus in the absence of failures the approach imposes no overhead.

Introducing Fault-Tolerant Group Membership into the Collaborative Computing Transport Layer

In this paper we introduce the novel election based fault tolerance mechanisms recently incorporated into the Collaborative Computing Transport Layer (CCTL). CCTL offers the atomic reliable multicast facilities used in the Collaborative Computing Framework (CCF). Our approach utilizes a reliable IP multicast primitive to implement two electorial algorithms that not only form consensus, but efficiently deliver a compact matrix based view of the network. This matrix can subsequently be analyzed to identify specific network failures (e.g. partitioning). The underlying premise of the approach being that by basing fault tolerance on a reliable multicast primitive, we eliminate the need for specific keep-alive packets such as heartbeats.

Paradigms for the parallel programming of heterogeneous machines through an interface compiler

The Reading Template Interface Compiler offers a series of communications frameworks to help with the structuring of applications written in C that need to control many parallel processes and processors. They include One to Many, All to All, Ring, Pipe, Grid and Toricgrid. They have been successfully applied to a variety of application codes, including the NAS kernels of the Parkbench suite and run over PVM (Parallel Virtual Machine).

Experimenting with local networks using an integrated circuit data link controller

Abstract There has been considerable discussion recently about the implementation of local networks. Much of this discussion centres around techniques requiring special hardware which is often too expensive for the application or not yet available. There is, however, an immediate need for local networks of a more modest price and performance. This paper describes the use of an integrated circuit data link controller to implement an inexpensive low performance local network.