A study of adaptive load balancing algorithms for distributed systems

With the advent of distributed computer systems with a largely transparent user interface, new questions have arisen regarding the management of such an environment by an operating system. One fertile area of research is that of load balancing, which attempts to improve system performance by redistributing the workload submitted to the system by the users. Early work in this field concentrated on static placement of computational objects to improve performance, given prior knowledge of process behaviour. More recently this has evolved into studying dynamic load balancing with process migration, thus allowing the system to adapt to varying loads. In this thesis, we describe a simulated system which facilitates experimentation with various load balancing algorithms. The system runs under UNIX and provides functions for user processes to communicate through software ports; processes reside on simulated homogeneous processors, connected by a user-specified topology, and a mechanism is included to allow migration of a process from one processor to another. We present the results of a study of adaptive load balancing algorithms, conducted using the aforementioned simulated system, under varying conditions; these results show the relative merits of different approaches to the load balancing problem, and we analyse the trade-offs between them. Following from this study, we present further novel modifications to suggested algorithms, and show their effects on system performance.

The relative performance of scalable load balancing algorithms in loosely-coupled distributed systems. Available in 2 volumes.

The computer systems of today are characterised by data and program control that are distributed functionally and geographically across a network. A major issue of concern in this environment is the operating system activity of resource management for different processors in the network. To ensure equity in load distribution and improved system performance, load balancing is often undertaken. The research conducted in this field so far, has been primarily concerned with a small set of algorithms operating on tightly-coupled distributed systems. More recent studies have investigated the performance of such algorithms in loosely-coupled architectures but using a small set of processors. This thesis describes a simulation model developed to study the behaviour and general performance characteristics of a range of dynamic load balancing algorithms. Further, the scalability of these algorithms are discussed and a range of regionalised load balancing algorithms developed. In particular, we examine the impact of network diameter and delay on the performance of such algorithms across a range of system workloads. The results produced seem to suggest that the performance of simple dynamic policies are scalable but lack the load stability of more complex global average algorithms.

Ensuring efficiency and robustness in MANET

This paper introduces a joint load balancing and hotspot mitigation protocol for mobile ad-hoc network (MANET) termed by us as 'load_energy balance + hotspot mitigation protocol (LEB+HM)'. We argue that although ad-hoc wireless networks have limited network resources - bandwidth and power, prone to frequent link/node failures and have high security risk; existing ad hoc routing protocols do not put emphasis on maintaining robust link/node, efficient use of network resources and on maintaining the security of the network. Typical route selection metrics used by existing ad hoc routing protocols are shortest hop, shortest delay, and loop avoidance. These routing philosophy have the tendency to cause traffic concentration on certain regions or nodes, leading to heavy contention, congestion and resource exhaustion which in turn may result in increased end-to-end delay, packet loss and faster battery power depletion, degrading the overall performance of the network. Also in most existing on-demand ad hoc routing protocols intermediate nodes are allowed to send route reply RREP to source in response to a route request RREQ. In such situation a malicious node can send a false optimal route to the source so that data packets sent will be directed to or through it, and tamper with them as wish. It is therefore desirable to adopt routing schemes which can dynamically disperse traffic load, able to detect and remove any possible bottlenecks and provide some form of security to the network. In this paper we propose a combine adaptive load_energy balancing and hotspot mitigation scheme that aims at evenly distributing network traffic load and energy, mitigate against any possible occurrence of hotspot and provide some form of security to the network. This combine approach is expected to yield high reliability, availability and robustness, that best suits any dynamic and scalable ad hoc network environment. Dynamic source routing (DSR) was use as our underlying protocol for the implementation of our algorithm. Simulation comparison of our protocol to that of original DSR shows that our protocol has reduced node/link failure, even distribution of battery energy, and better network service efficiency.

Improving scheduling techniques in heterogeneous systems with dynamic, on-line optimisations

Computational performance increasingly depends on parallelism, and many systems rely on heterogeneous resources such as GPUs and FPGAs to accelerate computationally intensive applications. However, implementations for such heterogeneous systems are often hand-crafted and optimised to one computation scenario, and it can be challenging to maintain high performance when application parameters change. In this paper, we demonstrate that machine learning can help to dynamically choose parameters for task scheduling and load-balancing based on changing characteristics of the incoming workload. We use a financial option pricing application as a case study. We propose a simulation of processing financial tasks on a heterogeneous system with GPUs and FPGAs, and show how dynamic, on-line optimisations could improve such a system. We compare on-line and batch processing algorithms, and we also consider cases with no dynamic optimisations.

Evolutionary market agents for resource allocation in decentralised systems

We introduce self-interested evolutionary market agents, which act on behalf of service providers in a large decentralised system, to adaptively price their resources over time. Our agents competitively co-evolve in the live market, driving it towards the Bertrand equilibrium, the non-cooperative Nash equilibrium, at which all sellers charge their reserve price and share the market equally. We demonstrate that this outcome results in even load-balancing between the service providers. Our contribution in this paper is twofold; the use of on-line competitive co-evolution of self-interested service providers to drive a decentralised market towards equilibrium, and a demonstration that load-balancing behaviour emerges under the assumptions we describe. Unlike previous studies on this topic, all our agents are entirely self-interested; no cooperation is assumed. This makes our problem a non-trivial and more realistic one.