Automatic network management strategies have become paramount for meeting the needs of innovative real-time and data-intensive applications, such as those in the Internet of Things. However, the ever-growing and fluctuating demands for data and services in such applications require more than ever an efficient, scalable, and energy-aware network resource management. To address these challenges, this paper introduces a novel approach that leverages a modular architecture based on serverless functions within an energy-aware environment. By deploying SDN services as Functions as a Service (FaaS), the proposed approach enables dynamic, on-demand network function deployment, achieving significant cost and energy savings through fine-grained resource provisioning. Unlike previous monolithic SDN approaches, this work disaggregates SDN control plane into modular, serverless components, transforming tightly integrated functionalities into independent, on-demand services while ensuring performance, scalability, and energy efficiency. An analytical model is presented to approximate the service delivery time and power consumption, as well as an open source prototype implementation supported by an extensive experimental evaluation. Experimental results demonstrate significant improvement in energy efficiency compared to traditional approaches, highlighting the potential of this approach for sustainable network environments.

The advent of 5G drives the growth of multi-access edge computing (MEC), a revolutionary paradigm that utilises edge resources to enable low-latency mobile access and support complex service execution. Deploying services across geographically distributed edge nodes challenges providers to optimise performance metrics like end-to-end latency and resource efficiency, impacting user experience, operational cost, and environmental footprint. The energy harvesting (EH) technology provides clean and renewable energy at the edge, promoting the MEC system to minimise the impacts on the environment. However, the integration of EH can introduce energy limits and uncertainty to the powered devices. In the context of service scheduling with data flow dependencies, we propose two offline and heuristic-based service placement algorithms that balance minimizing latency and maximizing resource efficiency with fast execution. The two algorithms, evaluated in a simulated environment using state-of-the-art workload benchmarks, achieve significant energy consumption improvements while maintaining comparable latency. Based on the designed algorithms, we take a step further by developing an online dynamic resource scheduling and service offloading approach for MEC systems with EH capabilities. Simulation results demonstrate that the proposed strategy effectively utilise the harvested energy while granting a low user-experienced latency and low operational cost.

6G's AI native vision of embedding advance intelligence in the network while bringing it closer to the user requires a systematic evaluation of Generative AI (GenAI) models on edge devices. Rapidly emerging solutions based on Open RAN (ORAN) and Network-in-aBox strongly advocate the use of low-cost, off-the-shelf components for simpler and efficient deployment, for example, in provisioning rural connectivity. In this context, conceptual architecture, hardware testbeds, and precise performance quantification of Large Language Models (LLMs) on off-theshelf edge devices remain largely unexplored. This research investigates computationally demanding LLM inference on a single commodity Raspberry Pi serving as an edge testbed for ORAN. We investigate various LLMs, including small, medium, and large models, on a Raspberry Pi 5 Cluster using a lightweight Kubernetes distribution (K3s) with modular prompting implementation. We study its feasibility and limitations by analyzing throughput, latency, accuracy, and efficiency. Our findings indicate that CPU-only deployment of lightweight models, such as Yi, Phi, and Llama3, can effectively support edge applications, achieving a generation throughput of 5 to 12 tokens per second with less than 50% CPU and RAM usage. We conclude that GenAI on the edge offers localized inference in remote or bandwidthconstrained environments in 6 G networks without reliance on cloud infrastructure.

The integration of machine learning (ML) techniques has catalyzed significant advancements in the realm of autonomous vehicle technology, particularly in the domain of Intelligent Transport Systems (ITS) and the evolution of Connected and Automated Vehicles (CAVs). This study focuses on a downlink communication network characterized by a single-antenna Base Transceiver Station (BTS) and autonomous vehicles, with the BTS transmitting information at varying power levels. The primary objective is to predict optimal transmit power for vehicles across diverse channel conditions using machine learning methodologies, aimed at mitigating interference within the system. This interdisciplinary research endeavors to optimize transmit power from the BTS to vehicles through the synergy of machine learning and optimization techniques. By addressing this imperative, we aim to enhance vehicle safety, efficiency, and reliability within modern transportation networks. Leveraging advanced ML models, including Long Short-Term Memory (LSTM) and Feedforward Neural Network (FNN), our investigation reveals promising insights into the efficacy of these algorithms in advancing autonomous driving technologies. The paper presents comparative analyses of two prominent machine learning models, with the Mean Square Error (MSE) computed at 17.2516 for LSTM and 13.8562 for the Feedforward Model. These results underscore the potential of ML-driven approaches in optimizing transmit power for autonomous vehicle communication networks, thereby contributing to the ongoing evolution of intelligent transportation systems.

Serverless computing is revolutionising cloud application development as it offers the ability to create modular,\nhighly-scalable, fault-tolerant applications, with minimal operational management. In order to contribute\nto its widespread adoption of serverless platforms, the design and performance of language runtimes that\nare available in Function-as-a-Service (FaaS) serverless platforms is key. This paper aims to investigate the\nperformance impact of language runtimes in open-source serverless platforms, deployable on local clusters.\nA suite of experiments is developed and deployed on two selected platforms: OpenWhisk and Fission. The\nresults show a clear distinction between compiled and dynamic languages in cold starts but a pretty close\noverall performance in warm starts. Comparisons with similar evaluations for commercial platforms reveal\nthat warm start performance is competitive for certain languages, while cold starts are lagging behind by a wide\nmargin. Overall, the evaluation yielded usable results in regards to preferable choice of language runtime for\neach platform

Efficient application scheduling is critical for achieving high performance in heterogeneous computing systems . This problem has proved to be NP-complete even for the homogeneous case , heading research efforts in obtaining low complexity heuristics that produce good quality schedules. Such an example is HEFT, one of the most efficient list scheduling heuristics in terms of makespan and robustness. In this paper, we propose two task scheduling methods for heterogeneous computing systems that can be integrated to several task scheduling algorithms . First, a method that improves the scheduling time (the time for obtaining the output schedule) of a family of task scheduling algorithms is delivered without sacrificing the schedule length, when the computation costs of the application tasks are unknown. Second, a method that improves the scheduling length (makespan) of several task scheduling algorithms is proposed, by identifying which tasks are going to be executed as single-threaded and which as multi-threaded implementations, as well as the number of the threads used. We showcase both methods by using HEFT popular algorithm, but they can be integrated to other algorithms too, such as HCPT, HPS, PETS and CPOP. The experimental results, which consider 14580 random synthetic graphs and five real world applications , show that by enhancing HEFT algorithm with the two proposed methods, significant makespan gains and high scheduling time gains, are achieved.

A Context-aware Prediction Framework (CAPF) can be provided through a Self-adaptive System (SAS) resource manager to support the autoscaling decision in Edge Computing (EC) environments. However, EC dynamicity and workload fluctuation represent the main challenges to design a robust prediction framework. Machine Learning (ML) algorithms show a promising accuracy in workload forecasting problems which may vary according to the workload pattern. Therefore, the accuracy of such algorithms needs to be evaluated and compared in order to select the most suitable algorithm for EC workload prediction. In this paper, a thorough comparison is conducted focusing on the most popular ML algorithms which are Linear Regression (LR), Support Vector Regression (SVR), and Neural Networks (NN) using real EC dataset. The experimental results show that a robust prediction framework can be supported by more than one algorithm considering the EC contextual behavior. The results also reveal that the NN outpe rforms LR and SVR in most cases.

Hardware in High Performance Computing environments in recent years have increasingly become more heterogeneous in order to improve computational performance. An additional aspect of such systems is the management of power and energy consumption. The increase in heterogeneity requires middleware and programming model abstractions to eliminate additional complexities that it brings, while also offering opportunities such as improved power management. In this paper, we explore application level self-adaptation including aspects such as automated configuration and deployment of applications to different heterogeneous infrastructure and for their redeployment. This therefore not only mitigates complexities associated with heterogeneous devices but aims to take advantage of the heterogeneity. The overall result of this paper is a self-adaptive framework that manages application Quality of Service (QoS) at runtime, which includes the automatic migration of applications between different accelerated infrastructures. Discussion covers when this migration is appropriate and quantifies the likely benefits.

Summary Energy consumption in Cloud and High Performance Computing platforms is a significant issue and affects aspects such as the cost of energy and the cooling of the data center. Host level monitoring and prediction provides the groundwork for improving energy efficiency through the placement of workloads. Monitoring must be fast and efficient without unnecessary overhead, to enable scalability. This precludes the use of Watt meters attached per host, requiring alternative approaches such as integrated measurements and models. IPMI and RAPL are subject to error and partial measurement, which may be mitigated. Models allow for prediction and more responsive measures of power consumption, but require calibrating. The causes of calibration error are discussed, along with mitigation strategies, without overly complicating the underlying model. An outcome is a Watt meter emulator that provides hosts level power measurement along with estimated power consumption for a given workload, with an average error of 0.20W.

The key to optimizing software is the correct choice, order as well parameters of optimizations-transformations, which has remained an open problem in compilation research for decades for various reasons. First, most of the compilation subproblems-transformations are interdependent and thus addressing them separately is not effective. Second, it is very hard to couple the transformation parameters to the processor architecture (e.g., cache size and associativity) and algorithm characteristics (e.g. data reuse); therefore compiler designers and researchers either do not take them into account at all or do it partly. Third, the search space (all different transformation parameters) is very large and thus searching is impractical.

Virtual Machines (VMs) live migration is one of the important approaches to improve resource utilisation and support energy efficiency in Clouds. However, VMs live migration leads to performance loss and additional costs due to increased migration time and energy overhead. This paper introduces a Performance and Energy-based Cost Prediction Framework to estimate the total cost of VMs live migration by considering the resource usage and power consumption, while maintaining the expected level of performance. A series of experiments conducted on a Cloud testbed show that this framework is capable of predicting the workload, power consumption and total cost for heterogeneous VMs before and after live migration, with the possibility of recovering the migration cost e.g. 28.48% for the predicted cost recovery of the VM.

Virtual Machines (VMs) auto-scaling is an important technique to provision additional resource capacity in a Cloud environment. It allows the VMs to dynamically increase or decrease the amount of resources as needed in order to meet Quality of Service (QoS) requirements. However, the auto-scaling mechanism can be time-consuming to initiate (e.g. in the order of a minute), which is unacceptable for VMs that need to scale up/out during the computation, besides additional costs due to the increase of the energy overhead. This paper introduces a Performance and Energy-based Cost Prediction Framework to estimate the total cost of VMs auto-scaling by considering the resource usage and power consumption, while maintaining the expected level of performance. A series of experiments conducted on a Cloud testbed show that this framework is capable of predicting the auto-scaling workload, power consumption and total cost for heterogeneous VMs, with a cost-saving of up to 25% for the predicted total cost of VM self-configuration as compared to the current approaches in literature.

Efficient application scheduling is critical for achieving high performance in heterogeneous computing systems. This problem has proved to be NP-complete, heading research efforts in obtaining low complexity heuristics that produce good quality schedules. Although this problem has been extensively studied in the past, all the related works assume the computation costs of application tasks on processors are available a priori, ignoring the fact that the time needed to run/simulate all these tasks is orders of magnitude higher than finding a good quality schedule, especially in heterogeneous systems. In this paper, we propose two new methods applicable to several task scheduling algorithms for heterogeneous computing systems. We showcase both methods by using HEFT well known and popular algorithm, but they are applicable to other algorithms too, such as HCPT, HPS, PETS and CPOP. First, we propose a methodology to reduce the scheduling time of HEFT when the computation costs are unknown, without sacrificing the length of the output schedule (monotonic computation costs); this is achieved by reducing the number of computation costs required by HEFT and as a consequence the number of simulations applied. Second, we give heuristics to find which tasks are going to be executed as Single-Thread and which as Multi-Thread CPU implementations, as well as the number of the threads used. The experimental results considering both random graphs and real world applications show that extending HEFT with the two proposed methods achieves better schedule lengths, while at the same time requires from 4.5 up to 24 less simulations.

Cloud computing providers resort to a variety of techniques to improve energy consumption at each level of the cloud computing stack. Most of these techniques consider resource-level energy optimization at IaaS layer. This paper argues energy gains can be obtained by creating a cooperation between the PaaS layer (in charge of hosting the application/service) and the IaaS layer (in charge of handling the computing resources). It presents a novel method based on steering information and decision taking to trigger the PaaS and IaaS layers to adapt their energy mode in service operation, therefore enabling the Cloud stack to actively adapt to changing situations. Experimental results demonstrate such adaptation achieves dynamic energy management in each of the PaaS and IaaS cloud layers.

Energy consumption in Cloud computing is a significant issue and affects aspects such as the cost of energy, cooling in the data center and the environmental impact of cloud data centers. Monitoring and prediction provides the groundwork for improving the energy efficiency of data centers. This monitoring however is required to be fast and efficient without unnecessary overhead. It is also required to scale to the size of a data center where measurement through directly attached Watt meters is unrealistic. This therefore requires models that translate resource utilisation into the power consumed by a physical host. These models require calibrating and are hence subject to error. We discuss the causes of error within these models, focusingupon the use of IPMI in order to gather this data. We make recommendations on ways to mitigate this error without overly complicating the underlying model. The final result of these models is a Watt meter emulator that can provide values for power consumption from hosts in the data center, with an average error of 0.20W.

Cloud service providers offer access to their resources through formal service level agreements (SLA), and need well-balanced infrastructures so that they can maximise the quality of service (QoS) they offer and minimise the number of SLA violations. This paper focuses on a specific aspect of risk assessment as applied in cloud computing: methods within a framework that can be used by cloud service providers and service consumers to assess risk during service deployment and operation. It describes the various stages in the service lifecycle whereas risk assessment takes place, and the corresponding risk models that have been designed and implemented. The impact of risk on architectural components, with special emphasis on holistic management support at service operation, is also described. The risk assessor is shown to be effective through the experimental evaluation of the implementation, and is already integrated in a cloud computing toolkit.

Summary Grids offer best effort services to users. Service level agreements offer the opportunity to provide guarantees upon services offered, in such a way that it captures the users' requirements, while also considering concerns of the service providers. This is achieved via a process of converging requirements and service cost values from both sides towards an agreement. This paper presents the intelligent scheduling for quality of service market‐oriented mechanism for brokering guarantees upon completion time and cost for jobs submitted to a batch‐oriented compute service. Web Services agreement (negotiation) is used along with the planning of schedules in determining pricing, ensuring that jobs become prioritised depending on their budget constraints. An evaluation is performed to demonstrate how market mechanisms can be used to achieve this, whilst also showing the effects that scheduling algorithms can have upon the market in terms of rescheduling. The evaluation is completed with a comparison of the broker's capabilities in relation to the literature. Copyright © 2014 John Wiley & Sons, Ltd.

The importance of Quality of Service management in service oriented environments has brought the need of QoS aware solutions. Proactive adaptation approaches enable composite services to detect in advance, according to their QoS values, the need for a change in order to prevent upcoming problems, and maintain the functional and quality levels of the composition. This paper presents a proactive adaptation mechanism that implements self-optimization based on fuzzy logic. The optimization model uses two fuzzy inference systems that evaluate the QoS values of composite services, based on historical and freshly collected data, and decide if adaptation is needed or not. Experimental results show significant improvements in the global QoS of the use case scenarios, providing reductions of up to 8.9% in response time and 14.7% in energy consumption, and an improvement of 41% in availability; this is achieved with an average increment in cost of 11.75 %.

Autonomic management within autonomic computing framework is considered as the future and viable solution for many appliances, either in software or hardware. Nevertheless, its current research application in computer networks is mainly visible in the intra domain space, and less attention is given to inter domain between one core network and another. This paper reviews some of the work on autonomic management and presents a framework that can be extended to a global and universal solution, such as fulfilling demand on bandwidth management, Quality of Service (QOS), and Service Level Agreements (SLA). The autonomic computing self- features are considered to show the viability of the proposed framework.

The anticipated introduction of new forms of intelligent transport systems (ITS) represents a significant opportunity for increased cooperative mobility and sociotechnical changes within the transport system. Although such technologies are currently technically feasible, various socioeconomic and environmental barriers impede their arrival. This paper uses a recently developed ITS performance assessment framework, i.e., Environmental Fusion (EnvFUSION), to perform dynamic forecasting of the performance for three key ITS technologies: active traffic management (ATM), intelligent speed adaptation (ISA), and an automated highway system (AHS) using a mathematical theory of evidence. A consequential lifecycle assessment (c-LCA) is undertaken, which forms part of a data fusion process using data from various sources. The models forecast improvements for the three ITS technologies in line with social acceptability, economic profitability, and major carbon reduction scenarios up to 2050 on one of the U.K.'s most congested highways. An analytical hierarchy process (AHP) and the Dempster-Shafer theory (DST) are used to weight criteria that form part of an intelligent transport sustainability index (ITSI). Overall performance is then synthesized. Results indicate that there will be a substantial increase in socioeconomic and emissions benefits, provided that the policies are in place and targets are reached, which would otherwise delay their realization.

In order to improve service execution in Clouds, the management of Cloud Infrastructure has to take measures to adhere to Service Level Agreements and Business Level Objectives, from the application layer through to how services are supported at the lowest hardware levels. In this paper a risk model methodology and holistic management approach is developed specific to the operation of the Cloud Infrastructure Provider and is applied through improvements to SLA fault tolerance in Cloud Infrastructure. Risk assessments are used to analyse execution specific data from the Cloud Infrastructure and linked to a business driven holistic management component that is part of a Cloud Manager. Initial results show improved eco-efficiency, virtual machine availability and reductions in SLA failure across the whole Cloud infrastructure by applying our combined risk-based fault tolerance approach.

Resource usage in Clouds can be improved by deploying applications with richer defined requirements. Such "richer requirements" involve wider application / user specific context capture expressed in interrelated models. The use of model based requirements is presented using input from test-beds monitoring resource use in terms of Trust, Risk, Eco-Efficiency and Cost (TREC) models. The results of this application illustrate the potential that richer requirements have for better management of resources in Clouds.

Service-Oriented Computing (SOC) provides a flexible framework in which applications may be built up from services, often distributed across a network. One of the promises of SOC is that of Dynamic Binding where abstract consumer requests are bound to concrete service instances at runtime, thereby offering a high level of flexibility and adaptability. Existing research has so far focused mostly on the design and implementation of dynamic binding operations and there is little research into a comprehensive evaluation of dynamic binding systems, especially in terms of system failure and dependability. In this paper, we present a novel, extensible evaluation framework that allows for the testing and assessment of a Dynamic Binding System (DBS). Based on a fault model specially built for DBS's, we are able to insert selectively the types of fault that would affect a DBS and observe its behavior. By treating the DBS as a black box and distributing the components of the evaluation framework we are not restricted to the implementing technologies of the DBS, nor do we need to be co-located in the same environment as the DBS under test. We present the results of a series of experiments, with a focus on the interactions between a real-life DBS and the services it employs. The results on the NECTISE Software Demonstrator (NSD) system show that our proposed method and testing framework is able to trigger abnormal behavior of the NSD due to interaction faults and generate important information for improving both dependability and performance of the system under test.

Cloud computing provides outsourcing of resources bringing economic benefits. The outsourcing however does not allow data owners to outsource the responsibility of confidentiality, integrity and access control, as it still is the responsibility of the data owner. As cloud computing is transparent to both the programmers and the users, it induces challenges that were not present in previous forms of distributed computing. Furthermore, cloud computing enables its users to abstract away from low-level configuration such as configuring IP addresses and routers. It creates an illusion that this entire configuration is automated. This illusion is also true for security services, for instance automating security policies and access control in cloud, so that individuals or end-users using the cloud only perform very high-level (business oriented) configuration. This paper investigates the security challenges posed by the transparency of distribution, abstraction of configuration and automation of services by performing a detailed threat analysis of cloud computing across its different deployment scenarios (private, bursting, federation or multi-clouds). This paper also presents a risk inventory which documents the security threats identified in terms of availability, integrity and confidentiality for cloud infrastructures in detail for future security risks. We also propose a methodology for performing security risk assessment for cloud computing architectures presenting some of the initial results.

Cloud transformations require dynamic redistribution of resources across cloud infrastructure. From a legal perspective this movement of data from one data processor to another without the explicit consent of the data subject is a threat to data privacy. Levels of assurance and accountability have to be provided from the cloud infrastructure providers to the data subject in order to maintain trust. In cases of Cloud Transformation multiple providers are present and passing accountability down the chain is essential. Existing Service Level Agreements (SLA) and policy based privacy implementations fail to provide the flexibility and accountability needed in establishing these new relationships. By introducing combined risk and privacy assessment alongside SLA negotiation, the legal and data management implications of Cloud Transformation events can be better accounted for. This will better protect the privacy of data subjects and increase confidence and trust in the Cloud computing platform.

We propose a cloud contextualization mechanism which operates in two stages, contextualization of VM images prior to service deployment (PaaS level) and self contextualization of VM instances created from the image (IaaS level). The contextualization tools are implemented as part of the OPTIMIS Toolkit, a set of software components for simplified management of cloud services and infrastructures. We present the architecture of our contextualization tools and the feasibility of our contextualization mechanism is demonstrated in a three tier web application scenario. Preliminary performance results suggest acceptable performance and scalability of our prototype.

In this paper, we investigate how cloud computing could benefit computational neuroscience. To that end, Multiple Interacting Instantiations of Neuronal Dynamics (MIIND), a computational neuroscience modelling toolkit, was ported to a private, university-owned cloud. The aim was to pave the way for making MIIND more accessible to non-specialist users in virtue of concealing its implementation context as well as rendering local IT infrastructure unnecessary. For that purpose, a customisable MIIND-based workflow model was encased within a virtualised wrapping apparatus. This served to fully automate running configurable MIIND simulations remotely via a convenient web-interface in a transparent manner with the user being incognisant of the cloud and the service orientated architecture behind it. This architecture can be adopted for any application conforming to the workflow characteristics of MIIND and helps inform the porting process of serial and legacy applications.

The principles of risk management have been introduced in grid computing to help document and anticipate certain risks and manage them to ensure job executions are successful. Clouds are more complex environments with further concerns like risk, trust, eco-efficiency, green, security or cost. In this paper we present ongoing research work to analyze and address the risk factor in clouds with the aim of optimizing cloud services. The main contribution of this work is the presentation of a methodology for performing risk assessment in cloud environments including the target use cases, risk identification, mitigation and monitoring. Together with the corresponding mitigation strategies, the methodology provides technological assurance that will lead to a high confidence of Cloud service consumers on one side, and a cost effective and reliable productivity of cloud Service/Infrastructure Providers on the other side. The design of the risk assessment framework and its software toolkit implementation are part of the research and development work of the OPTIMIS (Optimized Infrastructure Services) project whose objective is to enable an open and dependable Cloud Service Ecosystem that delivers IT services that are adaptable, reliable, auditable and sustainable both ecologically and economically. The paper presents some preliminary results on the risk assessment of a Service/Infrastructure Provider at the cloud service deployment stage.

As a technology, cloud computing has become an IT buzzword for the past few years. Cloud computing has often been used with synonymous terms such as software as a service, platform as a service, and infrastructure as a service (IaaS). Cloud computing has the potential to advance research discoveries by making data and computing resources readily available at an unprecedented economy of scale and with tremendous scalability. This paper discusses the importance of QoS and Iaas performance in cloud computing. The results of a quantitative evaluation are presented into the performance overheads of propagating virtual machine (VM) images to physical resources, at the Iaas layer and then accessing the images, via a Hypervisor's virtual block I/O device. Two virtual infrastructure managers are evaluated: Nimbus and OpenNebula, alongside two VM managers: XEN and KVM. Nimbus is found to outperform OpenNebula, while XEN outperforms KVM in the majority of cases. Conclusions are drawn from the results on the suitability of these technologies for data-intensive applications and applications requiring highly dynamic resource sets, where making an uninformed decision on what technology to use could prevent an application reaching its full potential, once deployed onto a cloud.

Abstract Service level agreements (SLAs) are facilitators for widening the commercial uptake of Grid technology. They provide explicit statements of expectation and obligation between service consumers and providers. However, without the ability to assess the probability that an SLA might fail, commercial uptake will be restricted, since neither party will be willing to agree. Therefore, risk assessment mechanisms are critical to increase confidence in Grid technology usage within the commercial sector. This paper presents an SLA brokering mechanism with risk assessment support, which evaluates the probability of SLA failure. WS‐Agreement and risk metrics are used to facilitate SLA creation between service consumers and providers within a typical Grid resource usage scenario. An evaluation is conducted to examine risk models, the performance of the broker's implementation as well as a comparison of its capabilities against similar SLA‐based solutions from the literature. Copyright © 2011 John Wiley & Sons, Ltd.

If Grid computing is to experience widespread commercial adoption, then incorporating risk assessment and management techniques is essential,both during negotiation between service provider and service requester and during run-time. This paper focuses on the role of a resource broker in this context. Specifically, an approach to evaluating the reliability of risk information received from resource providers is presented, using historical data to provide a statistical estimate of the average integrity of their risk assessments, with respect to systematic overestimation or underestimation of the probability of failure. Simulation results are presented, indicating the effectiveness of this approach.

Service Level Agreements (SLAs) are explicit statements about all expectations and obligations in the business partnership between customers and providers. They have been introduced in Grid computing to overcome the best effort approach, making the Grid more interesting for commercial applications. However, decisions on negotiation and system management still rely on static approaches, not reflecting the risk linked with decisions. The EC-funded project "AssessGrid" aims at introducing risk assessment and management as a novel decision paradigm into Grid computing. This paper gives a general motivation for risk management and presents the envisaged architecture of a "risk-aware" Grid middleware and Grid fabric, highlighting its functionality by means of three showcase scenarios.

Resource brokering is an essential component in building effective Grid systems. Existing mechanisms employ a traditional approach for resource allocation, which is likely to run into performance problems. This paper presents the development of a broker that is designed within the SNAP (Service Negotiation and Acquisition Protocol) framework and focuses on applications that require resources on demand. The broker uses a three-phase commit protocol as the traditional advance reservation facilities cannot cater to such needs due to the prior time that it requires to schedule the reservation. Experiments have been carried out on a Grid testbed, supported by mathematical modelling and simulation. The experimental results show that the inclusion of the three-phase commit protocol results in a performance enhancement in terms of the time taken from submission of user requirements until a job begins execution. The broker is a viable contender for use in future Grid resource broker implementations.

Summary form only given. Resource brokering is an essential component in building effective grid systems. Existing mechanisms employ a traditional approach for resource allocation, which is likely to run into performance problems. We present the development of a broker that is designed within the SNAP (service negotiation and acquisition protocol) framework and focuses on applications that require resources on demand. The broker uses a three-phase commit protocol, as the traditional advance reservation facilities cannot cater for such needs due to the prior time that it requires to schedule the reservation. The experimental results show that the inclusion of the three-phase commit protocol results in a performance enhancement in terms of the time taken from submission of user requirements until a job begins execution. The broker is a viable contender for use in future grid resource broker implementations.

Abstract The list of applications requiring high‐performance computing resources is constantly growing. The cost of inter‐processor communication is critical in determining the performance of massively parallel computing systems for many of these applications. This paper considers the feasibility of a commodity processor‐based system which uses a free‐space optical interconnect. A novel architecture, based on this technology, is presented. Analytical and simulation results based on an implementation of BSP (Bulk Synchronous Parallelism) are presented, indicating that a significant performance enhancement, over architectures using conventional interconnect technology, is possible. Copyright © 2004 John Wiley & Sons, Ltd.

A grid system must integrate heterogeneous resources with varying quality and availability. For example, the load on any given resource may increase during execution of a time-constrained job. This places importance on the system's ability to recognize the state of these resources. This paper presents an approach used as a basis for system adaptation in which grid jobs are maintained at runtime. A reflective technique is used to simplify the adaptation in the grid application. The design of an adaptable resource broker is described and experimentally evaluated. Reflection is incorporated into the broker to separate functional and non-functional aspects of the system and facilitate the implementation of non-functional properties such as job migration. Results indicate that this approach enhances the likelihood of timely job completion in a dynamic grid system.