Search

In this paper, a non‐negative matrix factorization feature expansion (NMFFE) approach was proposed to overcome the feature‐sparsity issue when expanding features of short‐text. First, we took the internal relationships of short texts and words into account when segmenting words from texts and constructing their relationship matrix. Second, we utilized the Dual regularization non‐negative matrix tri‐factorization (DNMTF) algorithm to obtain the words clustering indicator matrix, which was used to get the feature space by dimensionality reduction methods. Thirdly, words with close relationship were selected out from the feature space and added into the short‐text to solve the sparsity issue. The experimental results showed that the accuracy of short text classification of our NMFFE algorithm increased 25.77%, 10.89%, and 1.79% on three data sets: Web snippets, Twitter sports, and AGnews, respectively compared with the Word2Vec algorithm and Char‐CNN algorithm. It indicated that the NMFFE algorithm was better than the BOW algorithm and the Char‐CNN algorithm in terms of classification accuracy and algorithm robustness.

Literate computing environments, such as the Jupyter (i.e., Jupyter Notebooks, JupyterLab, and JupyterHub), have been widely used in scientific studies; they allow users to interactively develop scientific code, test algorithms, and describe the scientific narratives of the experiments in an integrated document. To scale up scientific analyses, many implemented Jupyter environment architectures encapsulate the whole Jupyter notebooks as reproducible units and autoscale them on dedicated remote infrastructures (e.g., highperformance computing and cloud computing environments). The existing solutions are still limited in many ways, e.g., 1) the workflow (or pipeline) is implicit in a notebook, and some steps can be generically used by different code and executed in parallel, but because of the tight cell structure, all steps in the Jupyter notebook have to be executed sequentially and lack of the flexibility of reusing the core code fragments, and 2) there are performance bottlenecks that need to improve the parallelism and scalability when handling extensive input data and complex computation. In this work, we focus on how to manage the workflow in a notebook seamlessly. We 1) encapsulate the reusable cells as RESTful services and containerize them as portal components, 2) provide a composition tool for describing workflow logic of those reusable components, and 3) automate the execution on remote cloud infrastructure. Empirically, we validate the solution's usability via a use case from the Ecology and Earth Science domain, illustrating the processing of massive Light Detection and Ranging (LiDAR) data. The demonstration and analysis show that our method is feasible, but that it needs further improvement, especially on integrating distributed workflow scheduling, automatic deployment, and execution to develop as a mature approach.

Blockchain technologies, e.g., Hyperledger Fabric and Sawtooth, have been evolving rapidly during past years and enable potential decentralised innovations in a substantial amount of business applications, e.g. crowd journalism, car-sharing and energy trading. The development of decentralised business applications has to face challenges in selecting suitable blockchain technologies, customising network protocols among distributed peers, and optimising system performance to meet application requirements. Also, manually testing and comparing those different technologies are time-consuming. Therefore, an effective tool is needed for profiling the performance characteristics of blockchain services in different cloud environments. In this paper, we present the Customisable Blockchain-as-a-Service Performance Profiler (CBProf), a tool we developed for automating blockchain deployment and performance profiling in cloud environments. We also provide the implementation and functionality demonstration of this tool.

The advent of programmable network switch ASICs and recent developments on other programmable data planes (NPUs, FPGAs) drive the renewed interest in network data plane programmability. The P4 language has emerged as a strong candidate to describe a protocol independent datapath pipeline. With its supported architectures, the P4 language provides an excellent way to define the packet processing and forwarding behavior, while leaving other networking components such as the traffic management engine, to non-programmable fixed function elements, based on the capabilities of most programmable devices. However, network flexibility is essential to meet the Quality of Service (QoS) requirements of traffic flows. Thus, enabling programmable control for fixed-function elements like traffic management is crucial. Towards that end we propose the use of virtual queues in the P4 pipeline, investigate the application of virtual queue-based traffic management, and portability of the approach using different P4 programmable targets. Specifically, we focus on virtual queue based Active Queue Management (AQM) for congestion policing and meeting the latency targets of distinct network slices. The solution is compared to P4 built-in functionality for bandwidth management using meters, proving also that the additional dimensions of control are achieved without compromising the processing complexity of the solution.

Incentives are usually introduced by the regulator entity (third-party), to promote cooperation in a market. The implementation of incentives is always costly and thus might fail to be enforced sustainably. This work aims at exploring the effects of incentives from an institutional perspective, while coping with the scenario where the third-party is part of the system but not composed by players. The evolutionary game theory (EGT) framework is applied to identify the incentives that lead to pure cooperation. In contrast to traditional EGT, this paper introduces an elimination mechanism that can reduce the market size. The incentives identified in the EGT analysis are further examined in simulation experiments which measure the market size, affluence and sustainability. The findings show: (1) light punishment leads to a reduction of the market size, yet heavier punishment is beneficial to the market size and wealth; (2) mixed incentives will generally lead to different wealth of the third party and of the participants. While under moderate strength, the wealth of both parties is the same and their overall wealth is maximal; (3) for sustainability, pure punishment (resp. reward) is sustainable (resp. unsustainable), the sustainability of mixed incentives depends on both their strength and agents’ rationality level.

There are many organizations interested in sharing data with others, and they can do this only if a multi-domain secure platform is available. Such platforms, often referred to as Digital Data Marketplaces (DDMs), require that all the transactions follow the pre-defined policies that are established by the participating parties i.e, domains. However, building a multi-domain network infrastructure in which each domain can manage its own connectivity while at the same time all of the transactions follow the sharing agreements is still a challenge. In this paper, we introduce a multi-domain containerized DDM that is built upon a P4-based network. It can handle the communication of multiple domains and guarantee that the operation of transactions is based on the pre-defined policies. We also studied the setup performance by defining a model which we demonstrated follows the real measurements, and we can use for decision making. The results also show the low overhead of using P4 switch in network setup time. In addition, we conducted a security evaluation which showed that our P4-based network setup is secure against most types of attacks.

Security is a top concern in digital infrastructure and there is a basic need to assess the level of security ensured for any given application. To accommodate this requirement, we propose a new risk assessment system. Our system identifies threats of an application workflow, computes the severity weights with the modified Microsoft STRIDE/DREAD model and estimates the final risk exposure after applying security countermeasures in the available digital infrastructures. This allows potential customers to rank these infrastructures in terms of security for their own specific use cases. We additionally present a method to validate the stability and resolution of our ranking system with respect to subjective choices of the DREAD model threat rating parameters. Our results show that our system is stable against unavoidable subjective choices of the DREAD model parameters for a specific use case, with a rank correlation higher than 0.93 and normalised mean square error lower than 0.05.

The paradigm of Network Function Virtualization (NFV) requires the underlying networks to be able to route traffic through dynamically deployed nodes. The research presented here demonstrates the feasibility of using segment routing (SR), a modern incarnation of the source routing paradigm, and suitable SDN controllers to accomplish the traffic steering goal. Specifically, we focused on IPv4 SR and built a proof of concept SRMPLS network using Juniper vQFX routers with custom VNFs. The Juniper NorthStar SDN controller was used to build SR- MPLS LSPs via the Path Computation Element Communication Protocol (PCEP). We validated the operation of our proof of concept in two scenarios: VNFs reinstantiation and service chain creation and support. In both cases traffic can be successfully steered through the functions, hence demonstrating for the first time the feasibility of adoption of the proposed technologies.

We study the correlation clustering problem using the quantum approximate optimization algorithm (QAOA) and qudits, which constitute a natural platform for such non-binary problems. Specifically, we consider a neutral atom quantum computer and propose a full stack approach for correlation clustering, including Hamiltonian formulation of the algorithm, analysis of its performance, identification of a suitable level structure for ${}^{87}{\rm Sr}$ and specific gate design. We show the qudit implementation is superior to the qubit encoding as quantified by the gate count. For single layer QAOA, we also prove (conjecture) a lower bound of $0.6367$ ($0.6699$) for the approximation ratio on 3-regular graphs. Our numerical studies evaluate the algorithm's performance by considering complete and Erd\H{o}s-R\'enyi graphs of up to 7 vertices and clusters. We find that in all cases the QAOA surpasses the Swamy bound $0.7666$ for the approximation ratio for QAOA depths $p \geq 2$. Finally, by analysing the effect of errors when solving complete graphs we find that their inclusion severely limits the algorithm's performance., Comment: 30+12 pages, 14 figures, accepted into Quantum

Advances in automation, communication, sensing and computation enable experimental scientific processes to generate data at increasingly great speeds and volumes. Research infrastructures are devised to take advantage of these data, providing advanced capabilities for acquisition, sharing, processing, and analysis; enabling advanced research and playing an ever-increasing role in the environmental and earth science research domain. The ENVRI community identified several recurring requirements in the development of environmental research infrastructures such as i) duplication of efforts to solve similar problems; ii) lack of standards to harmonise and accelerate development, and bring about interoperability; iii) a large number of data models and data information systems within the domain, and iv) a steep learning curve for integration complex research infrastructure systems. To address these challenges, the ENVRI community has developed and re-fined the Environmental Research Infrastructures Reference Model (ENVRI Reference Model or ENVRI RM), a modelling framework encoding this knowledge. The proposed modelling framework encompasses a language and a notation to describe the research domain, its systems and the requirements and challenges faced when implementing those systems. By adopting ENVRI RM as an integrative approach, the environmental research community can secure interoperability between infrastructures, enable reuse, share resources, experiences and common language, reduce unnecessary duplication of effort, and speed up the understanding of research infrastructure systems. This chapter provides a short introduction to the ENVRI RM.

By effectively virtualizing operating systems and encapsulating necessary runtime contexts of software components and services, container technologies can significantly improve portability and efficiency for distributed application deployment. It flexibly extends virtual machine based cloud (Infrastructure-as-a-Service) as a much lighter virtual environment (container cluster) for agile application management. However, existing container management systems are not capable of handling concurrent requests efficiently, particularly for the underlying clusters with heterogeneous machines and the requested containers with multi-resource demands. In this paper, we propose an Enhanced Container Scheduler (ECSched) for efficiently scheduling concurrent container requests on heterogeneous clusters with multi-resource constraints. We formulate the container scheduling problem as a minimum cost flow problem (MCFP), and represent the container requirements using a specific graph data structure (flow network). ECSched affords flexibility in constructing the flow network based on a batch of concurrent requests, and performs the MCFP algorithm to schedule the concurrent requests in an online manner. We evaluate ECSched in different testbed clusters, and measure the scheduling overhead with large-scale simulations. The experimental results show that ECSched outperforms state-of-the-art container schedulers in container performance and resource efficiency, and only introduces a small and acceptable scheduling overhead in large-scale clusters.

The ARTICONF project funded by the European Horizon 2020 program addresses issues of trust, time-criticality and democratisation for a new generation of federated infrastructure, to fulfil the privacy, robustness, and autonomy related promises critical in proprietary social media platforms. It aims to: (1) simplify the creation of open and agile social media ecosystem with trusted participation using a two stage permissioned blockchain; (2) automatically detect interest groups and communities using graph anonymization techniques for decentralised and tokenized decision-making and reasoning; (3) elastically autoscale time-critical social media applications through an adaptive orchestrated Cloud edge-based infrastructure meeting application runtime requirements; and (4) enhance monetary inclusion in collaborative models through cognition and knowledge supply chains. We summarize the initial envisaged architecture of the ARTICONF ecosystem, the industrial pilot use cases for validating it, and the planned innovations compared to related other European research projects.

Research infrastructures available for researchers in environmental and earth science are diverse and highly distributed; dedicated research infra-structures exist for atmospheric science, marine science, solid earth science, biodiversity research, and more. These infrastructures aggregate and curate key research datasets and provide consolidated data services for a target re-search community, but they also often overlap in scope and ambition, sharing data sources, sometimes even sites, using similar standards, and ultimately all contributing data that will be essential to addressing the societal challenges that face environmental research today. Thus, while their diversity poses a problem for open science and multidisciplinary research, their commonalities mean that they often face similar technical problems and consequently have common requirements when addressing the implementation of best practices in curation, cataloguing, identification and citation, and other related core topics for data science.In this chapter, we review the requirements gathering performed in the context of the cluster of European environmental and earth science research infrastructures participating in the ENVRI community, and survey the common challenges identified from that requirements gathering process.

To support the current trend of personalised medicine, a collaboration between different healthcare providers is increasingly vital. The main element is the ability to share data among all parties while abiding by a data sharing policy. The EPI (Enabling Personalised Intervention) project addresses the problem of personalised diagnosis by developing real-time monitoring services and digital health twins. The EPI services run over adaptive computing infrastructures which provide more flexibility to accommodate the different requests. This paper proposes the EPI framework to support these novel health services over programmable infrastructure. The framework works on aligning the parties’ ability to share data with the policy defined beforehand. We explain the approach by introducing the framework’s data sharing logic model. We define the formalism of the logic model to deduce feasible data movements between and possibly satisfy a data collaboration request. We reinforce the framework’s logic model by introducing the algorithms running on this federated system to simulate its workflow. We provide three healthcare use cases running on a typical EPI infrastructure. We evaluated our model according to three relevant parameters, performance, feasibility, and aggregation power, and we can conclude that our framework supports the required interoperability between the EPI partners.

The real-world complex networks, such as biological, transportation, biomedical, web, and social networks, are usually dynamic and change over time. The communities which reflect the substructures hidden in the networks usually overlap each other, and detecting overlapping communities in the dynamic complex networks is a challenging task. Prior researchers have applied multiobjective optimization method to the detection of dynamic overlapping communities and achieved some excellent results. However, in terms of multiobjective processing, the prior studies all adopt the decomposition method based on weight parameters, and different weight parameters or different parameter values can easily affect the community detection results which further results in the uneven distribution of the detected results in the target space. To solve the above problems, a hybrid algorithm, that is, Collaborative Particle Swarm multiobjective Optimization-based Dynamic Overlapping Community Detection (CPSO-DOCD) algorithm is proposed in this paper. First, to improve the diversity of particles, the encoding/decoding of the particle and the cross inheritance and the variation of particle are redefined first based on label propagation. In each network snapshot, multiple particle swarms are initialized based on Community Overlap Propagation Algorithm (COPRA) to generate particles with uniform distribution. Multiple different objective functions are optimized using multiple particle swarms respectively to avoid the incorrect selection of weight parameters. In addition, a reference-point-based is adopted in the particle selecting stage to solve the uneven distribution of detected results in the target space. Second, a node label entropy-based particle swarm algorithm is proposed to improve the accuracy of community detection of current network snapshots. Finally, when one snapshot switches to another over time, a migration strategy based on COPRA local-search and clique generation is utilized to adjust the prior community detection results, which enables the former results can be adapted to the new network snapshots. The experiments are implemented based on four dynamic networks which are Cit-HepPh, Cit-HepTh, Emailed-EU-core-temporal, and CollegeMsg. The hypervolume value of the overlapping community detection result obtained by CPSO-DOCD is 0.5%–2% higher than MDOA, MCMOEA, SLPAD, and iLCD. Furthermore, CPSO-DOCD also performed better than MDOA, MCMOEA, SLPAD, and iLCD on C-metric values, and CPSO-DOCD can approach approximately to the Pareto frontier.

Modern scientific research challenges require new technologies, integrated tools, reusable and complex experiments in distributed computing infrastructures. But above all, computing power for efficient data processing and analyzing. Containers technologies have emerged as a new paradigm to address such intensive scientific applications problems. Their easy deployment in a reasonable amount of time and the few required computational resource make them more suitable. Containers are considered light virtualization solutions. They enable performance isolation and flexible deployment of complex, parallel, and high-performance systems. Moreover, they gained popularity to modernize and migrate scientific applications in computing infrastructure management. Additionally, they reduce computational time processing. In this paper, we first give an overview of virtualization and containerization technologies. We discuss the taxonomies of containerization technologies of the literature, and then we provide a new one that covers and completes those proposed in the literature. We identify the most important application domains of containerization and their technological progress. Furthermore, we discuss the performance metrics used in most containerization techniques. Finally, we point out research gaps in the related aspects of containerization technology that require more research.

Dataset repositories publish a significant number of datasets continuously within the context of a variety of domains, such as biodiversity and oceanography. To conduct multidisciplinary research, scientists and practitioners must discover datasets from various disciplines unfamiliar with them. Well-known search engines, such as Google dataset and Mendeley data, try to support researchers with cross-domain dataset discovery based on their contents. However, as datasets typically contain scientific observations or collected data from service providers, their contextual information is limited. Accordingly, effective dataset indexing can be impossible to increase the Findability, Accessibility, Interoperability, and Reusability (FAIRness) based on their contextual information. This paper presents an indexing pipeline to extend contextual information of datasets based on their scientific domains by using topic modeling and a set of suggested rules and domain keywords (such as essential variables in environment science) based on domain experts’ suggestions. The pipeline relies on an open ecosystem, where dataset providers publish semantically enhanced metadata on their data repositories. We aggregate, normalize, and reconcile such metadata, providing a dataset search engine that enables research communities to find, access, integrate, and reuse datasets. We evaluated our approach on a manually created gold standard and a user study.

With the increasing adoption of intelligent transportation systems and the upcoming era of autonomous vehicles, vehicular services (such as remote driving, cooperative awareness, and hazard warning) will have to operate in an ever-changing and dynamic environment. Anticipating the dynamics of traffic flows on the roads is critical for these services and, therefore, it is of paramount importance to forecast how they will evolve over time. By predicting future events (such as traffic jams) and demands, vehicular services can take proactive actions to minimize Service Level Agreement (SLA) violations and reduce the risk of accidents. In this paper, we compare several techniques, including both traditional time-series and recent Machine Learning (ML)-based approaches, to forecast the traffic flow at different road segments in the city of Torino (Italy). Using the the most accurate forecasting technique, we propose n-max algorithm as a forecast-based scaling algorithm for vertical scaling of edge resources, comparing its benefits against state-of-the-art solutions for three distinct Vehicle-to-Network (V2N) services. Results show that the proposed scaling algorithm outperforms the state-of-the-art, reducing Service Level Objective (SLO) violations for remote driving and hazard warning services. Work partially funded by the EU H2020 5GROWTH Project (grant no. 856709) and H2020 collaborative Europe/Taiwan research project 5G-DIVE (grant no. 859881).

Many computational problems are subject to a quantum speed-up: one might find that a problem having an O(n³)-time or O(n²)-time classic algorithm can be solved by a known O(n^{1.5})-time or O(n)-time quantum algorithm. The question naturally arises: how much quantum speed-up is possible? The area of fine-grained complexity allows us to prove optimal lower-bounds on the complexity of various computational problems, based on the conjectured hardness of certain natural, well-studied problems. This theory has recently been extended to the quantum setting, in two independent papers by Buhrman, Patro and Speelman [Buhrman et al., 2021], and by Aaronson, Chia, Lin, Wang, and Zhang [Aaronson et al., 2020]. In this paper, we further extend the theory of fine-grained complexity to the quantum setting. A fundamental conjecture in the classical setting states that the 3SUM problem cannot be solved by (classical) algorithms in time O(n^{2-ε}), for any ε > 0. We formulate an analogous conjecture, the Quantum-3SUM-Conjecture, which states that there exist no sublinear O(n^{1-ε})-time quantum algorithms for the 3SUM problem. Based on the Quantum-3SUM-Conjecture, we show new lower-bounds on the time complexity of quantum algorithms for several computational problems. Most of our lower-bounds are optimal, in that they match known upper-bounds, and hence they imply tight limits on the quantum speedup that is possible for these problems. These results are proven by adapting to the quantum setting known classical fine-grained reductions from the 3SUM problem. This adaptation is not trivial, however, since the original classical reductions require pre-processing the input in various ways, e.g. by sorting it according to some order, and this pre-processing (provably) cannot be done in sublinear quantum time. We overcome this bottleneck by combining a quantum walk with a classical dynamic data-structure having a certain "history-independence" property. This type of construction has been used in the past to prove upper bounds, and here we use it for the first time as part of a reduction. This general proof strategy allows us to prove tight lower bounds on several computational-geometry problems, on Convolution-3SUM and on the 0-Edge-Weight-Triangle problem, conditional on the Quantum-3SUM-Conjecture. We believe this proof strategy will be useful in proving tight (conditional) lower-bounds, and limits on quantum speed-ups, for many other problems., LIPIcs, Vol. 215, 13th Innovations in Theoretical Computer Science Conference (ITCS 2022), pages 31:1-31:12

In the classical RAM, we have the following useful property. If we have an algorithm that uses M memory cells throughout its execution, and in addition is sparse, in the sense that, at any point in time, only m out of M cells will be non-zero, then we may "compress" it into another algorithm which uses only m log M memory and runs in almost the same time. We may do so by simulating the memory using either a hash table, or a self-balancing tree. We show an analogous result for quantum algorithms equipped with quantum random-access gates. If we have a quantum algorithm that runs in time T and uses M qubits, such that the state of the memory, at any time step, is supported on computational-basis vectors of Hamming weight at most m, then it can be simulated by another algorithm which uses only O(m log M) memory, and runs in time Õ(T). We show how this theorem can be used, in a black-box way, to simplify the presentation in several papers. Broadly speaking, when there exists a need for a space-efficient history-independent quantum data-structure, it is often possible to construct a space-inefficient, yet sparse, quantum data structure, and then appeal to our main theorem. This results in simpler and shorter arguments., LIPIcs, Vol. 232, 17th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC 2022), pages 10:1-10:19

Virtual Research Environments (VREs), also known as science gateways or virtual laboratories, assist researchers in data science by integrating tools for data discovery, data retrieval, workflow management and researcher collaboration, often coupled with a specific computing infrastructure. Recently, the push for better open data science has led to the creation of a variety of dedicated research infrastructures (RIs) that gather data and provide services to different research communities, all of which can be used independently of any specific VRE. There is therefore a need for generic VREs that can be coupled with the resources of many different RIs simultaneously, easily customised to the needs of specific communities. The resource metadata produced by these RIs rarely all adhere to any one standard or vocabulary however, making it difficult to search and discover resources independently of their providers without some translation into a common framework. Cross-RI search can be expedited by using mapping services that harvest RI-published metadata to build unified resource catalogues, but the development and operation of such services pose a number of challenges. In this paper, we discuss some of these challenges and look specifically at the VRE4EIC Metadata Portal, which uses X3ML mappings to build a single catalogue for describing data products and other resources provided by multiple RIs. The Metadata Portal was built in accordance to the e-VRE Reference Architecture, a microservice-based architecture for generic modular VREs, and uses the CERIF standard to structure its catalogued metadata. We consider the extent to which it addresses the challenges of cross-RI search, particularly in the environmental and earth science domain, and how it can be further augmented, for example to take advantage of linked vocabularies to provide more intelligent semantic search across multiple domains of discourse.

Virtual Research Environments (VREs) provide user-centric support in the lifecycle of research activities, e.g., discovering and accessing research assets, or composing and executing application workflows. A typical VRE is often implemented as an integrated environment, which includes a catalog of research assets, a workflow management system, a data management framework, and tools for enabling collaboration among users. Notebook environments, such as Jupyter, allow researchers to rapidly prototype scientific code and share their experiments as online accessible notebooks. Jupyter can support several popular languages that are used by data scientists, such as Python, R, and Julia. However, such notebook environments do not have seamless support for running heavy computations on remote infrastructure or finding and accessing software code inside notebooks. This paper investigates the gap between a notebook environment and a VRE and proposes an embedded VRE solution for the Jupyter environment called Notebook-as-a-VRE (NaaVRE). The NaaVRE solution provides functional components via a component marketplace and allows users to create a customized VRE on top of the Jupyter environment. From the VRE, a user can search research assets (data, software, and algorithms), compose workflows, manage the lifecycle of an experiment, and share the results among users in the community. We demonstrate how such a solution can enhance a legacy workflow that uses Light Detection and Ranging (LiDAR) data from country-wide airborne laser scanning surveys for deriving geospatial data products of ecosystem structure at high resolution over broad spatial extents. This enables users to scale out the processing of multi-terabyte LiDAR point clouds for ecological applications to more data sources in a distributed cloud environment., A revised version has been published in the journal software practice and experience

Network Functions Virtualization (NFV) is an industry effort to replace traditional hardware middleboxes with virtualized network functions (VNFs) running on general-build hardware platforms, enabling cost reduction, operational efficiency, and service agility. A Service Function Chain (SFC) constitutes an end-to-end network service, formed by chaining together VNFs in specific order. Infrastructure providers and cloud service providers try to optimally allocate computing and network resources to SFCs, in order to reduce costs and increase profit margins. The corresponding resource allocation problem, known as SFC embedding problem, is proven to be NP-hard. Traditionally the problem has been formulated as Mixed Integer Linear Program (MILP), assuming each SFC’s requirements are known a priori, while the embedding decision is based on a snapshot of the infrastructure’s load at request time. Reinforcement learning (RL) has been recently applied, showing promising results, specifically in dynamic environments, where such assumptions are considered unrealistic. However, standard RL techniques such as Q-learning might not be appropriate for addressing the problem at scale, as they are often ineffective for high-dimensional domains. On the other hand, Deep RL (DRL) algorithms can deal with high dimensional state spaces. In this paper, a Deep Q-Learning (DQL) approach is proposed to address the SFC resource allocation problem. The DQL agent utilizes a neural network for function approximation in Q-learning with experience replay learning. The simulations demonstrate that the new approach outperforms the linear programming approach. In addition, the DQL agent can perform SFC request admission control in real time.

Utilising programmable infrastructures is a promising approach to support secure data-sharing across healthcare domains of different capabilities in terms of network and security. The EPI1 (Enabling Personalised Interventions) framework automates the setup of the underlying infrastructure while considering different requirements communicated by the EPI components, such as logic area generator and policy management system.In our approach to dynamically provide collaborative environments, we containerise network functions (VFs), that are shipped out and instantiated at the edge of the network. We use container-based Virtual Network Function (VNFs) to accomplish fast deployment, high reusability, and low-performance overhead.Traffic interception and redirection through the chain of containerised network functions is a core feature of our framework. In this paper, we focus on the implementation and design of this tool for packet interception and redirection. We evaluate the performance of two approaches: an NGINX-based reverse proxy method and a SOCKS protocol-based method. We benchmark them to determine the overhead compared to a direct data-sharing session with no proxy. We conclude that the reverse proxy performs better in terms of overhead. Nonetheless, the SOCKS-based method works on a lower network level support all traffic types and offer a higher processing rate. We compare the methods according to other performance parameters. Subsequently, the choice of method will depend on the application performance requirements.

Container clusters play an increasingly important role in cloud computing for processing dynamic computing tasks. The resource manager (i.e., orchestrater) of the cluster automates the scheduling of the dynamic requests, effectively manages the resources’ utilization across distributing infrastructure resources. For many applications, the requests to the cluster are often with restricted deadlines. The scheduling of container clusters is often tricky, especially when the cluster’s size is large and the load of the requests is dynamically changing. Machine learningbased approaches such as reinforcement learning have attracted lots of research attention during the past years; However, those approaches suffer from low robustness when the requests in an operational environment are changing and different from the training data sets. This paper investigates this problem by quantifying the robustness and proposing meta-gradient reinforcement learning to improve the robustness of classical reinforcement learning-based approaches. The proposed approach can lead to better deadline guarantees and faster adaptation for timecritical task scheduling under dynamic environments. We then empirically test the benefits of our method using both real-world and synthetic data sets. The evaluation results show that the proposed method outperforms the compared RL methods in scheduling performance and robustness.

Demands for a more trustworthy Internet are constantly increasing, in particular to support emerging critical services such as intelligent urban transport systems and smart energy grids. Such cyber-physical systems require more insight into the properties of network operators (e.g., in terms of the security posture of their equipment) and more control over which network operators transport their data, thus going well beyond the traditional security paradigm which the Internet security community currently focuses on (confidentiality, availability, and integrity). In this work-in-progress paper we propose the UPIN framework, which aims to fulfill these new trust requirements. The framework advances the state-of-the-art by defining components needed to incorporate transparency, accountability, and controllability into the Internet or other types of inter-domain networks. The framework is based on our analysis of a smart grid use case to understand the specific needs of critical service providers and a literature study on existing technologies. We also discuss our ongoing work, and the demands and challenges of implementing and deploying the UPIN framework.

Copyright © 2021 The Author(s). The VECMA toolkit enables automated Verification, Validation and Uncertainty Quantification (VVUQ) for complex applications that can be deployed on emerging exascale platforms and provides support for software applications for any domain of interest. The toolkit has four main components including EasyVVUQ for VVUQ workflows, FabSim3 for automation and tool integration, MUSCLE3 for coupling multiscale models and QCG tools to execute application workflows on high performance computing (HPC). A more recent addition to the VECMAtk is EasySurrogate for various types of surrogate methods. In this paper, we present five tutorials from different application domains that apply these VECMAtk components to perform uncertainty quantification analysis, use surrogate models, couple multiscale models and execute sensitivity analysis on HPC. This paper aims to provide hands-on experience for practitioners aiming to test and contrast with their own applications. This work was supported by the VECMA project, which has received funding from the European Union Horizon 2020 research and innovation programme under grant agreement No. 800925. The development of MUSCLE3 and its respective description was supported by the Netherlands eScience Center and NWO under the e-MUSC project. The development of ISR3D was supported by the InSilc project and the In Silico World (ISW) project (European Union Horizon 2020 research and innovation programme grant agreements #777119 and #101016503 respectively). The calculations were performed in the Poznan Supercomputing and Networking Center.

Peak mitigation is of interest to power companies as peak periods may require the operator to over provision supply in order to meet the peak demand. Flattening the usage curve can result in cost savings, both for the power companies and the end users. Integration of renewable energy into the energy infrastructure presents an opportunity to use excess renewable generation to supplement supply and alleviate peaks. In addition, demand side management can shift the usage from peak to off-peak times and reduce the magnitude of peaks. In this work, we present a data driven approach for incentive-based peak mitigation. Understanding user energy profiles is an essential step in this process. We begin by analysing a popular energy research dataset published by the Ausgrid corporation. Extracting aggregated user energy behavior in temporal contexts and semantic linking and contextual clustering give us insight into consumption and rooftop solar generation patterns. We implement, and performance test a blockchain-based prosumer incentivization system. The smart contract logic is based on our analysis of the Ausgrid dataset. Our implementation is capable of supporting 792,540 customers with a reasonably low infrastructure footprint.

Purpose The purpose of this paper is to boost multidisciplinary research by the building of an integrated catalogue or research assets metadata. Such an integrated catalogue should enable researchers to solve problems or analyse phenomena that require a view across several scientific domains. Design/methodology/approach There are two main approaches for integrating metadata catalogues provided by different e-science research infrastructures (e-RIs): centralised and distributed. The authors decided to implement a central metadata catalogue that describes, provides access to and records actions on the assets of a number of e-RIs participating in the system. The authors chose the CERIF data model for description of assets available via the integrated catalogue. Analysis of popular metadata formats used in e-RIs has been conducted, and mappings between popular formats and the CERIF data model have been defined using an XML-based tool for description and automatic execution of mappings. Findings An integrated catalogue of research assets metadata has been created. Metadata from e-RIs supporting Dublin Core, ISO 19139, DCAT-AP, EPOS-DCAT-AP, OIL-E and CKAN formats can be integrated into the catalogue. Metadata are stored in CERIF RDF in the integrated catalogue. A web portal for searching this catalogue has been implemented. Research limitations/implications Only five formats are supported at this moment. However, description of mappings between other source formats and the target CERIF format can be defined in the future using the 3M tool, an XML-based tool for describing X3ML mappings that can then be automatically executed on XML metadata records. The approach and best practices described in this paper can thus be applied in future mappings between other metadata formats. Practical implications The integrated catalogue is a part of the eVRE prototype, which is a result of the VRE4EIC H2020 project. Social implications The integrated catalogue should boost the performance of multi-disciplinary research; thus it has the potential to enhance the practice of data science and so contribute to an increasingly knowledge-based society. Originality/value A novel approach for creation of the integrated catalogue has been defined and implemented. The approach includes definition of mappings between various formats. Defined mappings are effective and shareable.

The strong exponential-time hypothesis (SETH) is a commonly used conjecture in the field of complexity theory. It essentially states that determining whether a CNF formula is satisfiable can not be done faster than exhaustive search over all possible assignments. This hypothesis and its variants gave rise to a fruitful field of research, fine-grained complexity, obtaining (mostly tight) lower bounds for many problems in P whose unconditional lower bounds are very likely beyond current techniques. In this work, we introduce an extensive framework of Quantum Strong Exponential-Time Hypotheses, as quantum analogues to what SETH is for classical computation. Using the QSETH framework, we are able to translate quantum query lower bounds on black-box problems to conditional quantum time lower bounds for many problems in P. As an example, we provide a conditional quantum time lower bound of Ω(n^1.5) for the Longest Common Subsequence and Edit Distance problems. We also show that the n² SETH-based lower bound for a recent scheme for Proofs of Useful Work carries over to the quantum setting using our framework, maintaining a quadratic gap between verifier and prover. Lastly, we show that the assumptions in our framework can not be simplified further with relativizing proof techniques, as they are false in relativized worlds., LIPIcs, Vol. 187, 38th International Symposium on Theoretical Aspects of Computer Science (STACS 2021), pages 19:1-19:19

Networking offers an interesting standpoint to discuss aspects of responsible computing. The focus on a single high-level task (to provide secure and reliable end-to-end communications), and other characteristics such as resource/load distribution and decentralization of control, are fruitful properties to bring to the foreground fundamental dimensions which are generally overlooked in the responsible AI discourse: the pluralism of policies that may be in place, of expectations that may be taken into account, of their mechanisms of production and update, and their entrenchment in operationalization. Elaborating on a recent proposal on responsible Internet, the paper aims to provide a wider view of what is needed for networking to become a responsible type of computing, providing arguments for an architecture accounting for dynamic higher-level policies and expectation artefacts.

Proceedings of: Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit), 8-11 June 2021, Porto, Portugal. Network slicing is well-recognized as a core 5G technology to enable heterogeneous vertical services sharing the same infrastructure. In this context, the H2020 5Growth project extends baseline 5G management and orchestration platforms to manage the life-cycle of real end-to-end, reliable, and secure network slices with performance guarantees. In this paper, we present 5Growth's approaches to (i) attain isolation across network slices, (ii) provide secure interfaces towards third parties, and (iii) exploit AI/ML to achieve reliability through automated anomaly detection. In our quest towards validating full-fledged 5G pilots, we demonstrate our slicing mechanisms in PoCs that include interacting with ICT-17 infrastructure. This work has been partially supported by EC H2020 5GPPP 5Growth project (Grant 856709).

Data manipulation is often considered a serious problem in industrial applications as data tampering can lead to inaccurate financial reporting or even a corporate security crisis. A correct representation of company data is essential for the companies’ core business processes and is requested by governments and investors. However, the current solution, third-party auditing, is expensive and cannot be fully trusted. In this paper, we present the Data Warehouse Decentralized Integrity Model (DWH-DIM) to validate the integrity of the data warehouse and replace the current process. To address the challenge that the existing distributed integrity verification models cannot handle GDPR and are limited by scalability, our model uses a distributed file system to store attributes that can be used for the integrity verification task. The blockchain further confirms the authenticity of the files. Based on the proposed model, we present a detailed implementation of the DWH-DIM tool. The implementation is tested with a use case and several benchmarks. Experimental results demonstrate that our proposed model is feasible and meets the requirement for certificate warehouse data.

To reap its full benefits, 5G must evolve into a scalable decentralized architecture by exploiting intelligence ubiquitously and securely across different technologies, network layers, and segments. In this article, we propose end-to-end and ubiquitous secure machine learning (ML)-powered intent-based networking (IBN). The IBN framework is aware of its state and context to autonomously take proactive actions for service assurance. It is integrated in a zero-touch control and orchestration framework featuring an ML function orchestrator to manage ML pipelines. The objective is to create an elastic and dynamic infrastructure supporting per-domain and end-to-end network and services operation. The solution is supported by a radio access network and forwarding plane, and a cloud/edge virtualization infrastructure with ML acceleration. The resulting framework supports application-level resilience and intelligence through replication and elasticity. An illustrative intelligent application use case is presented. This work was partially supported by the AEI IBON (PID2020-114135RB-I00) project and by the ICREA Institution.

Social media applications are essential for next generation connectivity. Today, social media are centralized platforms with a single proprietary organization controlling the network and posing critical trust and governance issues over the created and propagated content. The ARTICONF project [ 1 ] funded by the European Union’s Horizon 2020 program researches a decentralized social media platform based on a novel set of trustworthy, resilient and globally sustainable tools that address privacy, robustness and autonomy-related promises that proprietary social media platforms have failed to deliver so far. This paper presents the ARTICONF approach to a car-sharing decentralized application (DApp) use case, as a new collaborative peer-to-peer model providing an alternative solution to private car ownership. We describe a prototype implementation of the car-sharing social media DApp and illustrate through real snapshots how the different ARTICONF tools support it in a simulated scenario.

In today’s clinical practice, magnetic resonance imaging (MRI) is routinely accelerated through subsampling of the associated Fourier domain. Currently, the construction of these subsampling strategies - known as experimental design - relies primarily on heuristics. We propose to learn experimental design strategies for accelerated MRI with policy gradient methods. Unexpectedly, our experiments show that a simple greedy approximation of the objective leads to solutions nearly on-par with the more general non-greedy approach. We offer a partial explanation for this phenomenon rooted in greater variance in the non-greedy objective's gradient estimates, and experimentally verify that this variance hampers non-greedy models in adapting their policies to individual MR images. We empirically show that this adaptivity is key to improving subsampling designs.

Finding optimal parameter configurations for tunable GPU kernels is a non-trivial exercise for large search spaces, even when automated. This poses an optimization task on a non-convex search space, using an expensive to evaluate function with unknown derivative. These characteristics make a good candidate for Bayesian Optimization, which has not been applied to this problem before. However, the application of Bayesian Optimization to this problem is challenging. We demonstrate how to deal with the rough, discrete, constrained search spaces, containing invalid configurations. We introduce a novel contextual variance exploration factor, as well as new acquisition functions with improved scalability, combined with an informed acquisition function selection mechanism. By comparing the performance of our Bayesian Optimization implementation on various test cases to the existing search strategies in Kernel Tuner, as well as other Bayesian Optimization implementations, we demonstrate that our search strategies generalize well and consistently outperform other search strategies by a wide margin., In context of the 2021 International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS) at SuperComputing 2021. Pending publication in IEEE Computer Society Technical Consortium on High Performance Computing (TCHPC)

Regardless of the context and rationale, running distributed applications on geographically dispersed IT resources often comes with various technical and organizational challenges. If not addressed appropriately, these challenges may impede development, and in turn, scientific and business innovation. We have developed the Brane framework to support implementers in addressing these challenges. Brane utilizes containerization to encapsulate functionalities as portable building blocks. Through programmability, application orchestration can be expressed using an intuitive domain-specific language. As a result, end-users with limited programming experience are empowered to compose applications by themselves, without having to deal with the underlying technical details. They can do this from user-friendly interactive notebooks. In this paper, we introduce Brane, describe its components and features, and validate the framework with an implementation of a real-world scientific use case.

Network slicing plays a key role in the 5G ecosystem for verticals to introduce new use cases in the industrial sector, i.e., Industry 4.0. However, a widely recognized challenge of network slicing is to provide traffic isolation and concurrently satisfy diverse performance requirements, e.g., bandwidth and latency. Such challenge becomes even more important when serving a large number of network traffic flows under a resource-limited condition between distributed sites, e.g., factory floor and remote office. In this work, we present the capability to retain these two goals at the same time, by applying the virtual queue notion over a priority queuing based pipeline in P4 switch over software-defined networks. To examine the effectiveness of our approach, a proof-of-concept is setup to serve different requests of Industry 4.0 use cases over a mixed data path, including P4 switch and Open vSwitch, for a large number of network flows. This work was supported in part by the EU H2020 5GROWTH Project under Grant 856709.

We have set up the first multi-domain P4programmable network in the Netherlands, consisting of six different sites interconnected by SURF's optical network. The purpose of the network is to experiment with novel inter-domain network functions and emerging internet architectures, in particular to improve the transparency, accountability, and controllability of the Internet and future networks. We provide an overview of the testbed's set-up, examples of how we have used it, and our lessons learned.

Unmanned Aerial Vehicles (UAVs) have been extensively utilized to provide wireless connectivity in rural and under-developed areas, enhance network capacity and provide support for peaks or unexpected surges in user demand, mainly due to their fast deployment, cost-efficiency and superior communication performance resulting from Line of Sight (LoS)-dominated wireless channels. In order to exploit the benefits of UAVs as base stations or relays in a mobile network, a major challenge is to determine the optimal UAV placement and relocation strategy with respect to the mobility and traffic patterns of the ground network nodes. Moreover, considering that the UAVs form a multi-hop aerial network, capacity and connectivity constraints have significant impacts on the end-to-end network performance. To this end, we formulate the joint UAV placement and routing problem as a Mixed Integer Linear Program (MILP) and propose an approximation that leads to a LP rounding algorithm and achieves a balance between time-complexity and optimality.

We present here an implementation to support function chaining and migration in the Network Function Virtualization (NFV) paradigm, using Segment Routing (SR) and Path Computation Element Communication Protocol (PCEP). The paradigm of NFV requires the underlying networks to be able to route traffic through dynamically deployed Virtual Network Functions (VNFs). Segment Routing is a modern incarnation of the source routing paradigm, which together with PCEP we can use to build network paths, steering the traffic through specific segments. The proof of concept we built uses SR-MPLS, focused on IPv4 SR, and consists of Juniper vQFX routers with custom VNFs. The VNFs are custom developed BPF programs that run on special VNF hosts, integrated into the SR-MPLS network. To build SRMPLS paths through the network, we use the Juniper NorthStar Software Defined Networking (SDN) controller and the PCEP protocol. We developed two scenarios to validate the functionality of our proof of concept: the migration of VNFs and the creation of VNF service chains. We demonstrate that in both scenarios we can successfully steer the traffic through the VNFs, showing the feasibility of using SR with PCEP to assist the deployment of VNFs.

In modern society, citizens aspire to get trusted and reliable digital services to authenticate theirs to payments. With the COVID-19 crisis, online shopping’s fast growth has led citizens to increase registration in different systems. The registration is typically done without any guarantee that the involved business entity is trusted and that private data is managed adequately, namely according to the General Data Protection Regulation (GDPR). There are cases where online business adopts a federated authentication mechanism based on the existing and extensively adopted service providers, e.g., Facebook, and Google. With the European authorities’ complacency, this de facto trend seems to contribute to a dangerous unregulated digital services model. While avoiding the centralization risks, a possible alternative is to pursue the concept of regulated and competing digital online shops or services offered under a single collaborative model across Europe. Citizens aspire to get simple mechanisms based on a single provider for authentication and pay anywhere, even with some associated costs. In this direction, we propose a model that considers regulated providers managing citizens’ access to any online business in Europe, avoiding, in this way, the spreading of personal data across (business) organizations, thus decreasing the risk of personal data leaks. A collaborative network is foreseen to logically tie committed regulating authorities, providers, and digital online service providers. The proposed approach is ground on our previous research on systems integration, collaborative network infrastructure, and unified mobility payment services. This position paper offers a digital strategy for citizens, designated by Digital Person Ecosystem (DPE), which relies on Collaborative Networks concepts and centered on public authority leadership.

Network slicing plays a key role in the 5G ecosystem for vertical industries to introduce new services. However, one widely-recognized challenge of network slicing is to provide traffic isolation and concurrently satisfy diverse performance requirements, e.g., bandwidth and latency. In this work, we showcase the capability to retain these two goals at the same time, via extending the 5Growth baseline architecture and designing a new data-plane pipeline, i.e., virtual queue, over the P4 switch. To demonstrate the effectiveness of our approach, a proof-of-concept is presented serving different service requests over a mixed data path, including P4 switches and Open vSwitches (OvSs).

Proceedings of: Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit), 8-11 June 2021, Porto, Portugal. This paper presents a data-driven approach leveraging AI/ML models to automate the service scaling operation and, in this way, meet the service requirements while minimizing the consumption of network, computing, and storage resources. This approach is integrated into the 5Growth service management software platform. In particular, a prototype was developed to demonstrate how the novel 5Growth AI/ML platform can be used in a closed-loop automation system to support the automated service scaling operation. Furthermore, a number of additional ML-based approaches are developed in the context of eMBB and C-V2N scenarios, which can be embedded into the system for handling more complex use cases. This work has been partially supported by EC H2020 5GPPP 5Growth project (Grant 856709).

The Internet of Things (IoT) and mobile techniques enable real-time sensing for urban computing systems. By recruiting only a small number of users to sense data from selected subareas (namely cells), Sparse Mobile Crowdsensing (MCS) emerges as an effective paradigm to reduce sensing costs for monitoring the overall status of a large-scale area. The current Sparse MCS solutions reduce the sensing subareas (by selecting the most informative cells) based on the assumption that each sample has the same cost, which is not always realistic in realworld, as the cost of sensing in a subarea can be diverse due to many factors, e.g. condition of the device, location, and routing distance. To address this issue, we proposed a new cell selection approach consisting of three steps (information modeling, cost estimation, and cost-quality beneficial cell selection) to further reduce the total costs and improve the task quality. Specifically, we discussed the properties of the optimization goals and modeled the cell selection problem as a solvable bi-objective optimization problem under certain assumptions and approximation. Then, we presented two selection strategies, i.e. Pareto Optimization Selection (POS) and Generalized Cost-Benefit Greedy (GCBGREEDY) Selection along with our proposed cell selection algorithm. Finally, the superiority of our cell selection approach is assessed through four real-life urban monitoring datasets (Parking, Flow, Traffic, and Humidity) and three cost maps (i.i.d with dynamic cost map, monotonic with dynamic cost map and spatial correlated cost map). Results show that our proposed selection strategies POS and GCB-GREEDY can save up to 15.2% and 15.02% sample costs and reduce the inference errors to a maximum of 16.8% (15.5%) compared to the baseline- Query by Committee (QBC) in a sensing cycle. The findings show important implications in Sparse Mobile Crowdsensing for urban context properties.

Recommender systems should be able to handle highly sparse training data that continues to change over time. Among the many solutions, Ant Colony Optimization, as a kind of optimization algorithm modeled on the actions of an ant colony, enjoys the favorable characteristic of being optimal, which has not been easily achieved by other kinds of algorithms. A recent work adopting genetic optimization proposes a collaborative filtering scheme: Ant Collaborative Filtering (ACF), which models the pheromone of ants for a recommender system in two ways: (1) use the pheromone exchange to model the ratings given by users with respect to items, (2) use the evaporation of existing pheromone to model the evolution of users&rsquo, preference change over time. This mechanism helps to identify the users and the items most related, even in the case of sparsity, and can capture the drift of user preferences over time. However, it reveals that many users share the same preference over items, which means it is not necessary to initialize each user with a unique type of pheromone, as was done with the ACF. Regarding the sparsity problem, this work takes one step further to improve the Ant Collaborative Filtering&rsquo, s performance by adding a clustering step in the initialization phase to reduce the dimension of the rate matrix, which leads to the results that K&lt, &lt, users, where K is the number of clusters, which stands for the maximum number of types of pheromone carried by all users. We call this revised version the Improved Ant Collaborative Filtering (IACF). Experiments are conducted on larger datasets, compared with the previous work, based on three typical recommender systems: (1) movie recommendations, (2) music recommendations, and (3) book recommendations. For movie recommendation, a larger dataset, MoviesLens 10M, was used, instead of MoviesLens 1M. For book recommendation and music recommendation, we used a new dataset that has a much larger size of samples from Douban and NetEase. The results illustrate that our IACF algorithm can better deal with practical recommendation scenarios that handle sparse dataset.

The In-Stent Restenosis 2D model is a full y coupled multiscale simulation of post-stenting tissue growth, in which the most costly submodel is the blood flow simulation. This paper presents uncertainty estimations of the response of this model, as obtained by both non-intrusive and semi-intrusive uncertainty quantification. A surrogate model based on Gaussian process regression for non-intrusive uncertainty quantification takes the whole model as a black-box and maps directly the three uncertain inputs to the quantity of interest, the neointimal area. The corresponding uncertain estimates matched the results from quasi-Monte Carlo simulations well. In the semi-intrusive uncertainty quantification, the most expensive submodel is replaced with a surrogate model. We developed a surrogate model for the blood flow simulation by using a convolutional neural network. The semi-intrusive method with the new surrogate model offered efficient estimates of uncertainty and sensitivity while keeping a relatively high accuracy. It outperformed the results obtained with earlier surrogate models. It also achieved the estimates comparable to the non-intrusive method with a similar efficiency. Presented results on uncertainty propagation with non-intrusive and semi-intrusive metamodelling methods allow us to draw some conclusions on the advantages and limitations of these methods.

Last decades, Cloud computing has made significant impacts on traditional applications to change their development and operation methods. We witnessed ever more newly-built Clouds and data centers. However, the centralized management mechanism of current Clouds lacks the dispersion to satisfy the requirements of emerging collaborative applications, including AI, IoT, and autopilot. On the other hand, the Edge computing stays at the conceptual and experimental stage. Most organizations construct their own Edge nodes to operate applications. An efficient and incentive mechanism is missing to motivate the Edge and micro Cloud resource providers to join and constitute a more generalized and decentralized ecosystem. To address this issue, we propose ALLSTAR, a blockchain based architecture for equally combining all the Cloud and Edge resources to be seamlessly leveraged by the application in the DevOps (development and operations) lifecycle. The ALLSTAR architecture is a systematic solution to realize the "Cloud+Edge" management and contributes to constructing the corresponding ALLSTAR ecosystem. This paper describes the overall architecture of ALLSTAR, the related key techniques, and detailed application DevOps processes as well as the new business model.

The next generation of supercomputers will break the exascale barrier. Soon we will have systems capable of at least one quintillion (billion billion) floating-point operations per second (10 18 FLOPS). Tremendous amounts of work have been invested into identifying and overcoming the challenges of the exascale era. In this work, we present an overview of these efforts and provide insight into the important trends, developments, and exciting research opportunities in exascale computing. We use a three-stage approach in which we (1) discuss various exascale landmark studies, (2) use data-driven techniques to analyze the large collection of related literature, and (3) discuss eight research areas in depth based on influential articles. Overall, we observe that great advancements have been made in tackling the two primary exascale challenges: energy efficiency and fault tolerance. However, as we look forward, we still foresee two major concerns: the lack of suitable programming tools and the growing gap between processor performance and data bandwidth (i.e., memory, storage, networks). Although we will certainly reach exascale soon, without additional research, these issues could potentially limit the applicability of exascale computing.