The demand for Privacy-Preserving Machine Learning (PPML) is growing, facing challenges in privacy balance, computational efficiency, and real-world feasibility, as traditional cloud approaches often suffer from high latency and resource limitations. Our paper introduces an innovative approach leveraging Function as a Service (FaaS) and edge computing to address these issues, significantly accelerating encrypted ML inference with strong privacy guarantees. Using Hybrid Homomorphic Encryption (HHE) and a distributed serverless architecture, we build a scalable solution that limits computational overhead and maximises resource utilisation. Offloading compute-intensive ML inference tasks to stateless functions, allocated on-demand at the edge, enables parallel processing, minimising latency and improving execution time. Evaluations on real-world medical datasets show substantial improvements over conventional methods, demonstrating feasible low-latency, high-efficiency PPML in distributed environments. Our findings highlight the potential of edge-driven FaaS architectures to bridge security and speed, paving the way for practical, real-time, privacy-preserving AI.

Managing complex distributed applications in the Cloud-Edge continuum, including deployment on diverse resources and runtime operations, presents significant challenges. Orchestrators play a key role by automating resource discovery, optimisation, deployment, and life-cycle management while ensuring system performance. This paper introduces Swarmchestrate, a decentralised, application-centric orchestration framework inspired by self-organising Swarms. Our initial findings, based on the implementation in a Cloud-Edge simulator, demonstrate Swarmchestrate's potential, offering insights into resource coordination and optimised allocation for scalable systems.

With the emergence of Internet of Things (IoT) devices collecting large amounts of data at the edges of the network, a new generation of hyper-distributed applications is emerging, spanning cloud, fog, and edge computing resources.The automated deployment and management of such applications requires orchestration tools that take a deployment descriptor (e.g.Kubernetes manifest, Helm chart or TOSCA) as input, and deploy and manage the execution of applications at run-time.While most deployment descriptors are prepared by a single person or organisation at one specific time, there are notable scenarios where such descriptors need to be created collaboratively by different roles or organisations, and at different times of the application's life cycle.An example of this scenario is the modular development of digital twins, composed of the basic building blocks of data, model and algorithm.Each of these building blocks can be created independently from each other, by different individuals or companies, at different times.The challenge here is to compose and build a deployment descriptor from these individual components automatically.This paper presents a novel solution to automate the collaborative composition and generation of deployment descriptors for distributed applications within the cloud-to-edge continuum.The implemented solution has been prototyped in over 25 industrial use cases within the DIGITbrain project, one of which is described in the paper as a representative example.

Abstract Science gateways have been widely utilized by a large number of user communities to simplify access to complex distributed computing infrastructures. While science gateways are still becoming increasingly popular and the number of user communities is growing, the fast and efficient creation of new science gateways and the flexibility to deploy these gateways on‐demand on heterogeneous computational resources, remain a challenge. Additionally, the increase in the number of users, especially with very different backgrounds, requires intuitive embedded e‐learning tools that support all stakeholders to find related learning material and to guide the learning process. This paper introduces a novel science gateway framework that addresses these challenges. The framework supports the creation, publication, selection, and deployment of cloud‐based reference architectures that can be automatically instantiated and executed even by nontechnical users. The framework also incorporates a knowledge repository exchange and learning module that provides embedded e‐learning support. To demonstrate the feasibility of the proposed solution, two scientific case studies are presented based on the requirements of the plasmasphere, ionosphere, and thermosphere research communities.

Summary Scholars worldwide leverage science gateways/virtual research environments (VREs) for a wide variety of research and education endeavors spanning diverse scientific fields. Evaluating the value of a given science gateway/VRE to its constituent community is critical in obtaining the financial and human resources necessary to sustain operations and increase adoption in the user community. In this article, we feature a variety of exemplar science gateways/VREs and detail how they define impact in terms of, for example, their purpose, operation principles, and size of user base. Further, the exemplars recognize that their science gateways/VREs will continuously evolve with technological advancements and standards in cloud computing platforms, web service architectures, data management tools and cybersecurity. Correspondingly, we present a number of technology advances that could be incorporated in next‐generation science gateways/VREs to enhance their scope and scale of their operations for greater success/impact. The exemplars are selected from owners of science gateways in the Science Gateways Community Institute (SGCI) clientele in the United States, and from the owners of VREs in the International Virtual Research Environment Interest Group (VRE‐IG) of the Research Data Alliance. Thus, community‐driven best practices and technology advances are compiled from diverse expert groups with an international perspective to envisage futuristic science gateway/VRE innovations.

Summary As cloud adoption increases, so do the number of available cloud service providers. Moving complex applications between clouds can be beneficial—or other times necessary—but achieving this so‐called cloud portability is rarely straightforward. This article presents the adoption of OASIS TOSCA, a standard in the declarative description of cloud applications, to encourage and facilitate cloud portability in MiCADO, an application‐level multi‐cloud orchestration and auto‐scaling framework. The interface to MiCADO is an Application Description Template, which draws from the TOSCA specification to describe an application in MiCADO. The generic design of these templates is presented and their applicability for achieving portability between different container and cloud environments is analysed and evaluated. A proof‐of‐concept where MiCADO serves as the deployment and execution engine for a Science Gateway in Sleep Healthcare is then described. In this proof‐of‐concept, MiCADO facilitates the deployment of a complex healthcare application, which is then moved from one cloud service provider to another with only minimal changes to the template which originally described it. This TOSCA‐based approach to templates in MiCADO encourages movement between clouds by making cloud portability more approachable.

Abstract Automated deployment and run-time management of microservices-based applications in cloud computing environments is relatively well studied with several mature solutions. However, managing such applications and tasks in the cloud-to-edge continuum is far from trivial, with no robust, production-level solutions currently available. This paper presents our first attempt to extend an application-level cloud orchestration framework called MiCADO to utilise edge and fog worker nodes. The paper illustrates how MiCADO-Edge can automatically deploy complex sets of interconnected microservices in such multi-layered cloud-to-edge environments. Additionally, it shows how monitoring information can be collected from such services and how complex, user- defined run-time management policies can be enforced on application components running at any layer of the architecture. The implemented solution is demonstrated and evaluated using two realistic case studies from the areas of video processing and secure healthcare data analysis.

In the big data era, creating self-managing scalable platforms for running big data applications is a fundamental task. Such self-managing and self-healing platforms involve a proper reaction to hardware (e.g., cluster nodes) and software (e.g., big data tools) failures, besides a dynamic resizing of the allocated resources based on overload and underload situations and scaling policies. The distributed and stateful nature of big data platforms (e.g., Hadoop-based cluster) makes the management of these platforms a challenging task. This paper aims to design and implement a scalable cloud native Hadoopbased big data platform using MiCADO, an open-source, and a highly customisable multi-cloud orchestration and auto-scaling framework for Docker containers, orchestrated by Kubernetes. The proposed MiCADO-based big data platform automates the deployment and enables an automatic horizontal scaling (in and out) of the underlying cloud infrastructure. The empirical evaluation of the MiCADO-based big data platform demonstrates how easy, efficient, and fast it is to deploy and undeploy Hadoop clusters of different sizes. Additionally, it shows how the platform can automatically be scaled based on user-defined policies (such as CPU-based scaling).

Abstract Typical cloud applications require high-level policy driven orchestration to achieve efficient resource utilisation and robust security to support different types of users and user scenarios. However, the efficient and secure utilisation of cloud resources to run applications is not trivial. Although there have been several efforts to support the coordinated deployment, and to a smaller extent the run-time orchestration of applications in the Cloud, no comprehensive solution has emerged until now that successfully leverages applications in an efficient, secure and seamless way. One of the major challenges is how to specify and manage Quality of Service (QoS) properties governing cloud applications. The solution to address these challenges could be a generic and pluggable framework that supports the optimal and secure deployment and run-time orchestration of applications in the Cloud. A specific aspect of such a cloud orchestration framework is the need to describe complex applications incorporating several services. These application descriptions must specify both the structure of the application and its QoS parameters, such as desired performance, economic viability and security. This paper proposes a cloud technology agnostic approach to application descriptions based on existing standards and describes how these application descriptions can be processed to manage applications in the Cloud.

Abstract Molecular docking and virtual screening experiments require large computational and data resources and high-level user interfaces in the form of science gateways. While science gateways supporting such experiments are relatively common, there is a clearly identified need to design and implement more complex environments for further analysis of docking results. This paper describes a generic framework and a related methodology that supports the efficient development of such environments. The framework is modular enabling the reuse of already existing components. The methodology, which proposes three techniques that the development team can use, is agile and encourages active participation of end-users. Based on the framework and methodology, two prototype implementations of science-gateway-based docking environments are presented and evaluated. The first system recommends a receptor-ligand pair for the next docking experiment, and the second filters docking results based on ligand properties.

There are many scientific and commercial applications that require the execution of a large number of independent jobs resulting in significant overall execution time. Therefore, such applications typically require distributed computing infrastructures and science gateways to run efficiently and to be easily accessible for end-users. Optimising the execution of such applications in a cloud computing environment by keeping resource utilisation at minimum but still completing the experiment by a set deadline has paramount importance. As container-based technologies are becoming more widespread, support for job-queuing and auto-scaling in such environments is becoming important. Current container management technologies, such as Docker Swarm or Kubernetes, while provide auto-scaling based on resource consumption, do not support job queuing and deadline-based execution policies directly. This paper presents JQueuer, a cloud-agnostic queuing system that supports the scheduling of a large number of jobs in containerised cloud environments. The paper also demonstrates how JQueuer, when integrated with a cloud application-level orchestrator and auto-scaling framework, called MiCADO, can be used to implement deadline-based execution policies. This novel technical solution provides an important step towards the cost-optimisation of batch processing and job submission applications. In order to test and prove the effectiveness of the solution, the paper presents experimental results when executing an agent-based simulation application using the open source REPAST simulation framework.

Computational fluid dynamics (CFD) is widely used in manufacturing and engineering from product design to testing. CFD requires intensive computational power and typically needs high-performance computing to reduce potentially long experimentation times. Dedicated high-performance computing systems are often expensive for small-to-medium enterprises (SMEs). Cloud computing claims to enable low-cost access to high-performance computing without the need for capital investment. The cloud-based simulation platform for manufacturing and engineering simulation platform aims to provide a flexible and easy to use cloud-based platform-as-a-service (PaaS) technology that can enable SMEs to realize the benefits of high-performance computing. Our platform incorporates workflow management and multicloud implementation across various cloud resources. Here, we present the components of our technology and experiences in using it to create a cloud-based version of the Transport phenomena Analysis Tool CFD software. Three case studies favourably compare the performance of a local cluster and two different clouds and demonstrate the viability of our cloud-based approach.

Simulation is used in industry to study a large variety of problems ranging from increasing the productivity of a manufacturing system to optimising the design of a wind turbine. However, some simulation models can be computationally demanding and some simulation projects require time consuming experimentation. High performance computing infrastructures such as clusters can be used to speed up the execution of large models or multiple experiments but at a cost that is often too much for Small and Medium-sized Enterprises (SMEs). Cloud computing presents an attractive, lower cost alternative. However, developing a cloud-based simulation application can again be costly for an SME due to training and development needs, especially if software vendors need to use resources of different heterogeneous clouds to avoid being locked-in to one particular cloud provider. In an attempt to reduce the cost of development of commercial cloud-based simulations, the CloudSME Simulation Platform (CSSP) has been developed as a generic approach that combines an AppCenter with the workflow of the WS-PGRADE/gUSE science gateway framework and the multi-cloud-based capabilities of the CloudBroker Platform. The paper presents the CSSP and two representative case studies from distinctly different areas that illustrate how commercial multi-cloud-based simulations can be created.

Cloud computing offers massive scalability and elasticity required by many scientific and commercial applications. Combining the computational and data handling capabilities of clouds with parallel processing also has the potential to tackle Big Data problems efficiently. Science gateway frameworks and workflow systems enable application developers to implement complex applications and make these available for end-users via simple graphical user interfaces. The integration of such frameworks with Big Data processing tools on the cloud opens new opportunities for application developers. This paper investigates how workflow systems and science gateways can be extended with Big Data processing capabilities. A generic approach based on infrastructure aware workflows is suggested and a proof of concept is implemented based on the WS-PGRADE/gUSE science gateway framework and its integration with the Hadoop parallel data processing solution based on the MapReduce paradigm in the cloud. The provided analysis demonstrates that the methods described to integrate Big Data processing with workflows and science gateways work well in different cloud infrastructures and application scenarios, and can be used to create massively parallel applications for scientific analysis of Big Data.

This special issue represents an active collaboration between the organizers of two science gateway workshops.The International Workshop on Science Gateways 2013 has taken place in June 2013 in Zurich and the Science Gateway Institute Workshop 2013, held in conjunction with IEEE Cluster in September 2013 in Indianapolis.The workshops attracted together over 100 international researchers and have led to excellent presentations and publications on science gateway developments, workflow-centered enhancements, science gateway infrastructures, and developments in specific research domains such as the life sciences and health applications.Authors of accepted submissions to the workshops have been invited to submit extended versions for a special issue.This special issue consists of the accepted papers of a further peer review process.The increasing complexity of scientific study and the increasingly digital nature of data have resulted in a myriad of community-developed solutions.Advanced Web portals, also called science gateways, have emerged in many domains.They assemble the computational resources, data collections, visualization capabilities, collaboration tools, and even access to instruments that scientists need to conduct their research.Development of these gateways is also increasingly complex, and developers often find it quite valuable to learn from one another, even across domains.This special issue highlights accepted papers from two workshops.The Fifth annual International Workshop on Science Gateways, held June 2013 in Zurich and the Science Gateway Institute workshop, held in conjunction with IEEE Cluster held September 2013 in Indianapolis.Both workshops incorporated a peer review process.Authors of top papers from both events were invited to submit extended versions of their work for publication in this special issue.The purpose of these workshops is to provide a forum to showcase science gateway projects and related technologies.Developers can learn from one another and learn about new technologies, and principal investigators can keep abreast on the state of the field.This special issue features contributions in the areas of technologies for building gateways, workflows to enhance the capabilities of gateways, and infrastructures that support science gateways.Also featured are ready-to-use gateways in the life sciences and health applications fields.Providers of science gateway technologies aim at offering generic frameworks to ease the development of science gateways for a specific research domain while providers of distributed computing infrastructures work on supporting communities with computing and data resources.Such infrastructures require policies on usage and security, and the close collaboration with developers of domain-specific science gateways elucidates the demands in the specific research domain.Users of such a domain want to focus on their research questions and create and analyze data in an intuitive and efficient way-regardless of whether the underlying infrastructure provides resources in cloud, grid, or cluster infrastructures. TECHNOLOGIES FOR SCIENCE GATEWAY DEVELOPMENTThe development of science gateways can be distinguished in two main tasks.Firstly, the generic part, which is concerned with security features, accesses to underlying infrastructures and job, workflow, and data management.This part can be very similar for diverse science gateways.Frameworks or APIs support developers with building blocks so that there is not the need to develop such features

In this paper we give methods for time-aware music recommendation in a social media service with the potential of exploiting immediate temporal influences between users. We consider events when a user listens to an artist the first time and this event follows some friend listening to the same artist short time before. We train a blend of matrix factorization methods that model the relation of the influencer, the influenced and the artist, both the individual factor decompositions and their weight learned by variants of stochastic gradient descent (SGD). Special care is taken since events of influence form a subset of the positive implicit feedback data and hence we have to cope with two different definitions of the positive and negative implicit training data. In addition, in the time-aware setting we have to use online learning and evaluation methods. While SGD can easily be trained online, evaluation is cumbersome by traditional measures since we will have potentially different top recommendations at different times. Our experiments are carried over the two-year "scrobble" history of 70,000 Last.fm users and show a 5% increase in recommendation quality by predicting temporal influences.

SUMMARY Molecular docking simulations have high potential to contribute to a wide area of molecular and biomedical research in various disciplines including molecular biology, drug design, environmental studies and psychology. Conducting large‐scale molecular docking experiments requires a vast amount of computing resources. Several types of distributed computing infrastructures have been investigated and utilized recently to conduct such simulations, including service and desktop grid systems or local clusters. This paper investigates and analyses how Windows Azure‐based cloud resources can be applied for this purpose. A virtual screening experiment framework has been implemented on a Windows Azure‐based cloud using the generic worker concept. Virtual machines can be instantiated in the cloud on demand scaling up the simulations based on the volume of molecules to be docked and the available financial resources. Bioscientists are able to execute the simulations and visualise the results from a high‐level user interface. The paper describes the experiences when implementing the molecular docking application on this novel platform and provides the first benchmarking experiments to evaluate the suitability of the infrastructure for computation intensive simulations. Copyright © 2013 John Wiley & Sons, Ltd.

Scientific workflows are common in biomedical research, particularly for molecular docking simulations such as those used in drug discovery. Such workflows typically involve data distribution between computationally demanding stages which are usually mapped onto large scale compute resources. Volunteer or Desktop Grid (DG) computing can provide such infrastructure but has limitations resulting from the heterogeneous nature of the compute nodes. These constraints mean that reducing the make span of a given workflow stage submitted to a DG becomes problematic. Late jobs can significantly affect the make span, often completing long after the bulk of the computation has finished. In this paper we present a system capable of significantly reducing the make span of a scientific workflow. Our system comprises a DG which is dynamically augmented with an infrastructure as a service (IaaS) Cloud. Using this solution, the Cloud resources are used to process replicated late jobs. Our system comprises a core component termed the scheduler, which implements an algorithm to perform late job detection, Cloud resource management (instantiation and reuse), and job monitoring. We offer a formal definition of this algorithm, whilst we also provide an evaluation of our prototype using a production scientific workflow.

Most vertebrates are able to make detours and find shortcuts to achieve economical navigation. This ability requires the animal to keep track its direction and distance from specific locations. In rodents, direction of the animal is coded by the activity of head direction cells present in several regions of the brain, but distance information is only indirectly available, through the entorhinal cortical grid cell system. A neural system downstream from the entorhinal cortex seems to be necessary to extract the distance information from the periodic activity of grid cells. We propose that a system of such cells store the distance of the animal from important locations in the dentate gyrus region of the hippocampus and these “distance cells” might be identified with the dentate granule cells. A computational model is set up to study the neural mechanism of distance information decoding from the ensemble of grids cells. The proposed distance cells receive innervation from entorhinal grid cells, the connection strength between grid cells and distance cells is set by a one-shot-learning rule and the distance cell activity is affected by a winner-take-all mechanism. Simulation results of this model verifies that the activity of the distance cell population is able to unambiguously code the distance of the animal from important places. The proposed distance cells have a multi-peaked, patchy spatial activity pattern similar to the firing pattern of granule cells in dentate gyrus.

A fundamental question in understanding neuronal computations is how dendritic events influence the output of the neuron. Different forms of integration of neighbouring and distributed synaptic inputs, isolated dendritic spikes and local regulation of synaptic efficacy suggest that individual dendritic branches may function as independent computational subunits. In the present paper, we study how these local computations influence the output of the neuron. Using a simple cascade model, we demonstrate that triggering somatic firing by a relatively small dendritic branch requires the amplification of local events by dendritic spiking and synaptic plasticity. The moderately branching dendritic tree of granule cells seems optimal for this computation since larger dendritic trees favor local plasticity by isolating dendritic compartments, while reliable detection of individual dendritic spikes in the soma requires a low branch number. Finally, we demonstrate that these parallel dendritic computations could contribute to the generation of multiple independent place fields of hippocampal granule cells.

The lack of widely accepted standards and the use of different middleware solutions divide today's Grid resources into non-interoperable production Grid islands. On the other hand, more and more experiments require such a large number of resources that the interoperation of existing production Grids becomes inevitable. This paper, based on the current results of grid interoperation studies, defines generic requirements towards the workflow level interoperation of grid solutions. It concentrates on intra-workflow interoperation of grid data resources, as one of the key areas of generic interoperation, and describes through an example how existing tools can be extended to achieve the required level of interoperation.

Manual deployment of the application usually requires expertise both about the underlying system and the application. Automatic service deployment can improve deployment significantly by using on-demand deployment and self-healing services. To support these features this paper describes an extension the globus workspace service. This extension includes creating virtual appliances for grid services, service deployment from a repository, and influencing the service schedules by altering execution planning services, candidate set generators or information systems.

So far BOINC based desktop grid systems have been applied at the global computing level. This paper describes an extended version of BOINC called SZTAKI desktop grid (SZDG) that aims at using desktop grids (DGs) at local (enterprise/institution) level. The novelty of SZDG is that it enables the hierarchical organisation of local DGs, i.e., clients of a DG can be DGs at a lower level that can take work units from their higher level DG server. More than that, even clusters can be connected at the client level and hence work units can contain complete MPI programs to be run on the client clusters. In order to easily create master/worker type DG applications a new API, called as the DC-API has been developed. SZDG and DC-API has been successfully applied both at the global and local level, both in academic institutions and in companies to solve problems requiring large computing power.

This article introduces a biologically realistic mathematical, computational model of theta (≈5 Hz) rhythm generation in the hippocampal CA1 region and some of its possible further applications in drug discovery and in robotic/computational models of navigation. The model shown here uses the conductance-based description of nerve cells: Populations of basket cells, alveus/lacunosum-moleculare interneurons, and pyramidal cells are used to model the hippocampal CA1 and a fast-spiking GABAergic interneuron population for modeling the septal influence. Results of the model show that the septo-hippocampal feedback loop is capable of robust theta rhythm generation due to proper timing of pyramidal cells and synchronization within the basket cell network via recurrent connections. © 2006 Wiley Periodicals, Inc. Int J Int Syst 21: 903–917, 2006.