Extreme learning machines is a neural network type that has been utilized in tasks such as regression and classification, due to their efficient training process, which is based on pseudoinverse matrices and randomized weights, avoiding the computationally intensive backpropagation. In order to further improve their performance and reduce their complexity with respect to number of required hyperparameters, especially in the case of multiple layer architectures, a novel multilayer adaptive approach, based on residual networks, is proposed. This approach constructs the network iteratively with respect to error minimization and parsimony using a recursive pseudoinverse matrix framework. A new block approach, using mixed precision arithmetic and Graphics Processing Units (GPU) is proposed. The proposed technique is coupled with a new adaptive penalty criterion to ensure adequate numbers of neurons are included in each layer, while avoiding highly correlated basis. Adaptive regularization, along with scaling, is also incorporated to ensure Symmetric Positive Definiteness (SPD) of the Gram matrix. Several random number distributions for the proposed approach are examined and discussed. Handling of large datasets is discussed and a new batch variant is proposed. The proposed scheme is evaluated for regression and classification tasks in a multitude of datasets and is compared with other neural network architectures.

Time series modelling is of significance to several scientific fields. Several approaches based on statistics, machine learning or combinations have been utilized. In order to model and forecast time series a novel parallel framework based on recursive pseudoinverse matrices is proposed. This framework enables the design of arbitrary statistical and machine learning models, adaptively, from a set of potential basis functions. This unification enables compact definition of existing and new models as well as easy implementation for new massively parallel architectures. The choice of appropriate basis functions is analysed and the fitting accuracy, termination criteria and model update operations are presented. A block variant for multivariate time series is also proposed. Parallel GPU implementation and performance optimization of the framework are provided, based on mixed precision arithmetic and matrix operations. The use of different basis functions is showcased with respect to various model univariate and multivariate time series for applications such as regression, frequency estimation and automatic trend detection. Discussions on limitations and future directions of research are also provided.

Modelling of large scale data series is of significant importance in fields such as astrophysics and finance. The continuous increase in available data requires new computational approaches such as the use of multicore processors and accelerators. Recently, a novel time series modelling and forecasting method was proposed, based on a recursively updated pseudoinverse matrix which enhances parsimony by enabling assessment of basis functions, before inclusion into the final model. Herewith, a novel GPU (Graphics Processing Unit) accelerated matrix based auto-regressive variant is presented, which utilizes lagged versions of a time series and interactions between them to form a model. The original approach is reviewed and a matrix multiplication based variant is proposed. The GPU accelerated and hybrid parallel versions are introduced, utilizing single and mixed precision arithmetic to increase GPU performance. Discussions around performance improvement and high order interactions are given. A block processing approach is also introduced to reduce memory requirements for the accelerator. Furthermore, the inclusion of constraints in the computation of weights, corresponding to the basis functions, with respect to the parallel implementation are discussed. The approach is assessed in a series of model problems and discussions are provided.

Time series modelling has a wide spectrum of applications in several fields including engineering and finance. Most traditional modelling techniques rely on assumptions related to the input data and manual pre-processing, based on user observations, rendering them unsuitable for analysing time-series with varying characteristics automatically, while more general modelling techniques usually require increased computational work for application and tuning. Recently, a general modelling framework based on a recursive Schur complement technique, that utilizes an adaptively determined set of basis functions, has been proposed. Herewith, a novel modified approach based on recursive incomplete pseudoinverse matrices in conjunction with preconditioned iterative methods for large datasets, is proposed. This sparse approach greatly reduces storage requirements and the recursive nature of the procedure avoids recomputation of the preconditioner after addition of a new basis function. Moreover, update of the coefficients for a different window of data and predefined basis functions can be performed utilizing the incomplete pseudoinverse matrix as preconditioner. The case of sinusoidal basis functions is presented along with a novel adaptive frequency estimation technique. The stability of the resulting model is discussed with respect to the choice of basis functions. The case of basis derived from machine learning techniques is also discussed. Numerical results are given depicting the applicability, generality and effectiveness of the proposed technique. Comparative results with other methods show forecasting RMSE improvement between 7% to 80%, for the majority of the chosen time series.

Time series modelling and forecasting techniques have a wide spectrum of applications in several fields including economics, finance, engineering and computer science. Most available modelling and forecasting techniques are applicable to a specific underlying phenomenon and its properties and lack generality of application, while more general forecasting techniques require substantial computational time for training and application. Herewith, we present a general modelling framework based on a recursive Schur - complement technique, that utilizes a set of basis functions, either linear or non-linear, to form a model for a general time series. The basis functions need not be orthogonal and their number is determined adaptively based on fitting accuracy. Moreover, no assumptions are required for the input data. The coefficients for the basis functions are computed using a recursive pseudoinverse matrix, thus they can be recomputed for different input data. The case of sinusoidal basis functions is presented. Discussions around stability of the resulting model and choice of basis functions is also provided. Numerical results depicting the applicability and effectiveness of the proposed technique are given.

With the increasing popularity of cloud computing, a new technology and business model called cloud service brokerage (CSB) is emerging. CSB is, in essence, a middleman in the cloud-computing supply chain to connect prospective cloud buyers with suitable service providers. This chapter focuses on a type of CSB, B2B cloud marketplaces. Recently, this type of marketplace has evolved into two broad categories—business application marketplaces and API marketplaces. This chapter reviews the characteristics of B2B cloud marketplaces, and their benefits, which include ease-of-use and ease-of-integration, enhanced security, increased manageability, faster implementation, and cost reduction. The chapter concludes with two mini-case studies, on Salesforce AppExchange and RapidAPI, to illustrate how firms could use B2B cloud marketplaces to generate, capture and measure business value.

SummaryFinding an appropriate resource to host the next application to be deployed in a Cloud environment can be a nontrivial task. To deliver the appropriate level of service, the functional requirements of the application must be met. Ideally, this process involves filtering the best resource from a number of possible candidates while simultaneously satisfying multiple objectives. If timely responses to resource requests are to be maintained, the sophistication of the filtering mechanism and size of the search space have to be carefully balanced. The quality of the solution will thus not readily scale with growth in cloud resources and filtering complexity. This limitation is becoming more evident with the emergence of hyperscale clouds and the increased complexity needed to accommodate the growing heterogeneity in resources. Moreover, meeting nonfunctional requirements, reflecting the Cloud Service Provider's business objects, is also becoming increasingly critical as service utilization and energy efficiency in a typical cloud deployment are extremely low. This paper proposes a re‐examination of the resource allocation problem by proposing a framework to support distributed resource allocation decisions and that can be dynamically populated with strategies to reflect the ever‐growing number of diverse objectives as they become evident in the evolving cloud infrastructure.

The lack of transparency surrounding cloud service provision makes it difficult for consumers to make knowledge based purchasing decisions. As a result, consumer trust has become a major impediment to cloud computing adoption. Cloud Trust Labels represent a means of communicating relevant service and security information to potential customers on the cloud service provided, thereby facilitating informed decision making. This research investigates the potential of a Cloud Trust Label system to overcome the trust barrier. Specifically, it examines the impact of a Cloud Trust Label on consumer perceptions of a service and cloud service provider trustworthiness and trust in the cloud service and cloud service provider. An experimental study was carried out with a sample of 227 business decision makers with data collected before exposure to the label to examine initial perceptions and after exposure to the label to examine any change in perceptions and attitudes. As hypothesised, the results suggest that Cloud Trust Labels that contain positive information can have a positive impact on trust and trustworthiness while Cloud Trust Labels that contain negative information have a negative impact. The practical implications of this new method of communicating trustworthiness online are discussed and recommendations are made for future research.

Within the complex context of high performance computing (HPC), the factors influencing technology adoption decisions remain largely unexplored. This study extends Diffusion of Innovation (DOI) and Human-Organization-Technology fit (HOT-fit) theories into an integrated model, to explore the impact of ten factors on cloud computing adoption decisions in the HPC context. The results suggest that adopters and non-adopters have different perceptions of the indirect benefits, adequacy of resources, top management support, and compatibility of adopting cloud computing for HPC. In addition, perceptions of the indirect benefits and HPC competences can be used to predict the cloud computing adoption decision for HPC. This is one of the first studies in the information systems (IS) literature exploring the factors impacting the cloud computing adoption decision in the important context of HPC. It integrates two influential technology adoption theories and enhances understanding of the key factors influencing organizations’ cloud computing adoption decisions in this context.

The ever-increasing number of customers that have been using cloud computing environments is driving heterogeneity in the cloud infrastructures. The incorporation of heterogeneous resources to traditional homogeneous infrastructures is supported by specific resource managers cohabiting with traditional resource managers. This blend of resource managers raises interoperability issues in the Cloud management domain as customer services are exposed to disjoint mechanisms and incompatibilities between APIs and interfaces. In addition, deploying and configuring HPC workloads in such environments makes porting HPC applications, from traditional cluster environments to the Cloud, complex and ineffectual. Many efforts have been taken to create solutions and standards for ameliorating interoperability issues in inter-cloud and multi-cloud environments and parallels exist between these efforts and the current drive for the adoption of heterogeneity in the Cloud. The work described in this paper attempts to exploit these parallels; managing interoperability issues in Cloud from a unified perspective. In this paper the mOSAIC ontology, pillar of the IEEE 2302 Standard for Intercloud Interoperability and Federation, is extended towards creating the CloudLightning Ontology (CL-Ontology), in which the incorporation of heterogeneous resources and HPC environments in the Cloud are considered. To support the CL Ontology, a generic architecture is presented as a driver to manage heterogeneity in the Cloud and, as a use case example of the proposed architecture, the internal architecture of the CloudLightning system is redesigned and presented to show the feasibility of incorporating a semantic engine to alleviate interoperability issues to facilitate the incorporation of HPC in Cloud. (C) 2018 Elsevier B.V. All rights reserved.

The addition of heterogeneous resources to conventional homogeneous cloud environments has enabled clouds to embrace a wide variety of new applications that heretofore were traditionally confined to specialized computing environments. The enhanced and extended features offered by heterogeneous resources enable service offerings that pose challenges to traditional cloud management throughout the entire service delivery stack. The accelerated uptake of heterogeneous resources is exacerbating these challenges, which no longer can be efficiently addressed in an ad-hoc manner. Therefore, an integrated approach to heterogeneous resource management that is cognizant of the unique advantages of different hardware types is needed. In this paper, two candidate approaches, a platform-integration scheme and a server-integration scheme, are introduced to address this management challenge. The platform-integration scheme integrates and coordinates the management of various coexisting resource managers and associated environments each of which may be managing resources of different types using the most appropriate resource abstraction method. In contrast, the server-integration scheme provides a single, lower level, fine-grained management mechanism across all hardware resource types. Ultimately, the goal of each schemes is to provide a unified view of resources from a capability perspective to consumers.

An overview of the traditional three-layer cloud architecture is presented as background for motivating the transition to clouds containing heterogeneous resources. Whereas this transition adds many important features to the cloud, including improved service delivery and reduced energy consumption, it also results in a number of challenges associated with the efficient management of these new and diverse resources. The CloudLightning architecture is proposed as a candidate for addressing this emerging complexity, and a description of its components and their relationships is given.

Traditionally, access to high performance computing was restricted by architectural complexity, availability of trained personnel, and budgetary issues. At the same time, research suggests that existing measures for greater data centre energy efficiencies will reach theoretical and practical limits in the near future. This concluding chapter briefly discusses the potential of (i) cloud computing to disrupt the high performance computing sector, and (ii) new heterogeneous cloud architectures, based on the concepts of self-organisation, self-management, and the separation of concerns, to disrupt extant cloud resource management approaches.

It gives us great pleasure to edit this special section on "Towards High Performance Computing in the Cloud," which contains papers from a number of high-profile authors in the field.These contributions were submitted via an open call for papers.All papers were rigorously reviewed and chosen on the basis of excellence and on the insight which they afford to the subject.As a result of the call for papers, twenty (20) articles were submitted and six (6) contributions were selected, resulting in an acceptance rate of 30%.HPC workloads have unique requirements in terms of performance, resource requirements and storage.However, several issues, such as interconnection speeds, data locality, placement of VMs and partial support for specialized hardware, limit the effectiveness of HPC applications in the Cloud.Nevertheless, in spite of ongoing challenges, recent years have seen HPC workloads migrating to the Cloud.This migration is driven by the desires to leverage novel

Driven by the successful business model, cloud computing is evolving rapidly from a moderate size data center consisting of homogeneous resources to a hyper-scale heterogeneous computing environment. The evolution has made the computing environment ever-increasingly complex, thus, raises challenges for the traditional approaches for managing a cloud environment in an efficient and effective manner. In response, a decentralized system architecture for cloud management is introduced. In this architecture, the management responsibility and resource organization in a conventional cloud environment are re-considered. The re-consideration results in composing a cloud environment into three entities including the Infrastructure, the Cloud Utility and Information Base, and Application Autonomous Systems. In this configuration, service providers focus on providing connected physical resources and introducing featured resources. Information related to the Infrastructure is stored and periodically updated in the Information Base. A consumer employs an Application Autonomous System for managing the life-cycle of a cloud application. An Application Autonomous System in the context of this paper is defined as a self-contained entity that encapsulates a cloud application, the associated resources and the management functions. An Application Autonomous System uses the Information Base and Cloud Utilities to locate and acquire desired resources, subsequently resources are deployed on the Infrastructure by invoking Cloud Utilities. Thereafter, the Application Autonomous System manages the life-cycle of both the application and the associated resources. Consumers are offered opportunities to employ preferred algorithms and strategies for this management. Thus, the responsibility of cloud application management and partially the resource management has shifted from service providers to the consumers in this decentralized system architecture.

A novel, general framework that can be used for constructing a self-organising and self-managing system is introduced. This framework is independent of the application domain. It embodies directed evolution, can be parameterised with different strategies, and supports both local and global goals. This framework is then used to apply the principles of self-organisation and self-management to resource management within the CloudLightning architecture.

Given its current development trajectory, the complexity of cloud computing ecosystems are evolving to where traditional resource management strategies will struggle to remain fit for purpose. These strategies have to cope with ever-increasing numbers of heterogeneous resources, a proliferation of new services, and a growing user-base with diverse and specialized requirements. This growth not only significantly increases the number of parameters needed to make good decisions, it increases the time needed to take these decisions. Consequently, traditional resource management systems are increasingly prone to poor decisions making. Devolving resources management decisions to the local environment of that resource can dramatically increase the speed of decisions making; moreover, the cost of gathering global information can thus be eliminated; saving communication costs. Experimental data, provided in this paper, illustrate that extant cloud deployments can be used as effective vehicles for devolved decision making. This finding strengthens the case for the proposed paradigm shift, since it does not require a change to the architecture of existing cloud systems. This shift would result in systems in which resources decide for themselves how best they can be used. This paper takes this idea to its logical conclusion and proposes a system for supporting self-managing resources in cloud environments. It introduces the concept of coalitions, consisting of collaborating resources, formed for the purpose of service delivery. It suggests the utility of restricting the interactions between the end-user and the cloud service provider to a well-defined services interface. It shows how clouds can be considered functionally, as engines for delivering an appropriate set of resources in response to service requests. And finally, since modern applications are increasingly constructed from sophisticated workflows of complex components, it shows how combinatorial auctions can be used to effectively deliver packages of resources to support those workflows.