CLOSER 2023 Abstracts

Area 1 - Cloud Computing Fundamentals
Full Papers Short Papers
Area 2 - Cloud Operations
Short Papers
Area 3 - Data as a Service
Short Papers
Area 4 - Edge Cloud and Fog Computing
Full Papers Short Papers
Area 5 - Service Modelling and Analytics
Short Papers
Area 6 - Services Science
Full Papers Short Papers
Area 7 - Cloud Computing Platforms and Applications
Full Papers Short Papers
Area 8 - Cloud Computing Enabling Technology
Full Papers Short Papers

Area 1 - Cloud Computing Fundamentals

Full Papers
Paper Nr: 15
Title:

Towards Poisoning of Federated Support Vector Machines with Data Poisoning Attacks
Authors:

Israt J. Mouri, Muhammad Ridowan and Muhammad A. Adnan
Abstract: Federated Support Vector Machine (F-SVM) is a technology that enables distributed edge devices to collectively learn a common SVM model without sharing data samples. Instead, edge devices submit local updates to the global machine, which are then aggregated and sent back to edge devices. Due to the distributed nature of federated learning, edge devices are vulnerable to poisoning attacks, especially during training. Attackers in adversarial edge devices can poison the dataset to hamper the global machine’s accuracy. This study investigates the impact of data poisoning attacks on federated SVM classifiers. In particular, we adopt two widespread data poisoning attacks for SVM named label flipping and optimal poisoning attacks for F-SVM and evaluate their impact on the MNIST and CIFAR10 datasets. We measure the impact of these poisoning attacks on the precision of global training. Results show that 33% of adversarial edge devices can reduce accuracy up to 30%. Furthermore, we also investigate some basic defense strategies against poisoning attacks on federated SVM.
Download
Short Papers
Paper Nr: 41
Title:

Latency-Aware Cost-Efficient Provisioning of Composite Applications in Multi-Provider Clouds
Authors:

Daniel C. Temp, Igor F. Capeletti, Ariel Goes de Castro, Paulo S. Severo de Souza, Arthur F. Lorenzon, Marcelo C. Luizelli and Fábio D. Rossi
Abstract: Composite applications have the versatility of maintaining several functions and services geographically distributed but being part of the same application. This particular software architecture fits in very easily with the model of distributed regions present in most cloud players. In this way, the search for leasing applications at the lowest cost becomes a reality, given that the application services can be in different players at a lower cost, as long as the performance metrics of the application as a whole are met. Performing provisioning decisions considering the allocation cost of different providers and the latency requirements of applications is not trivial, as these requirements are often conflicting, and finding good trade-offs involves the analysis of a large-scale combinatorial problem. Accordingly, this paper presents Clover, a placement algorithm that employs score-based heuristic procedures to find the best provisioning plan for hosting composite applications in geographically distributed cloud environments. Simulated experiments using real latency traces from Amazon Web Services indicate that Clover can achieve near-optimal results, reducing latency issues and allocation cost by up to 74.47% and 21.2%, respectively, compared to baseline strategies.
Download

Area 2 - Cloud Operations

Short Papers
Paper Nr: 25
Title:

Network Failures in Cloud Management Platforms: A Study on OpenStack
Authors:

Hassan M. Khan, Frederico Cerveira, Tiago Cruz and Henrique Madeira
Abstract: Cloud Management Platforms (CMPs) have a critical role in supporting private and public cloud computing as a tool to manage, provision and track resources and their usage. These platforms, like cloud computing, tend to be complex distributed systems spread across multiple nodes, thus network faults are a threat that can lead to failures to provide the expected service. This paper studies how network faults occurring in the links between the nodes of the CMP can propagate and affect the applications that are hosted on the virtual machines (VMs). We used fault injection to emulate various types of network faults in two links of the OpenStack CMP while a common cloud computing workload was being executed. The results show that not all network links have the same importance and that network faults can propagate and cause the performance of applications to degrade up to 50% and a small percentage of their operations to fail. Furthermore, in many campaigns some of the responses returned by the applications did not match the expected values.
Download
Paper Nr: 38
Title:

Analyzing Declarative Deployment Code with Large Language Models
Authors:

Giacomo Lanciano, Manuel Stein, Volker Hilt and Tommaso Cucinotta
Abstract: In the cloud-native era, developers have at their disposal an unprecedented landscape of services to build scalable distributed systems. The DevOps paradigm emerged as a response to the increasing necessity of better automations, capable of dealing with the complexity of modern cloud systems. For instance, Infrastructure-as-Code tools provide a declarative way to define, track, and automate changes to the infrastructure underlying a cloud application. Assuring the quality of this part of a code base is of utmost importance. However, learning to produce robust deployment specifications is not an easy feat, and for the domain experts it is time-consuming to conduct code-reviews and transfer the appropriate knowledge to novice members of the team. Given the abundance of data generated throughout the DevOps cycle, machine learning (ML) techniques seem a promising way to tackle this problem. In this work, we propose an approach based on Large Language Models to analyze declarative deployment code and automatically provide QA-related recommendations to developers, such that they can benefit of established best practices and design patterns. We developed a prototype of our proposed ML pipeline, and empirically evaluated our approach on a collection of Kubernetes manifests exported from a repository of internal projects at Nokia Bell Labs.
Download

Area 3 - Data as a Service

Short Papers
Paper Nr: 10
Title:

SkyData: Rise of the Data How Can the Intelligent and Autonomous Data Paradigm Become Real?
Authors:

Etienne Mauffret, Elise Jeanneau and Eddy Caron
Abstract: With the rise of Data as a Service, companies understood that whoever controls the data has the power. The past few years have exposed some of the weakenesses of traditional data management systems. For example, application owner can collect and use data to their own advantage without the user’s consent. We introduce in this paper the SkyData concept, which revolves around autonomous data evolving in a distributed system. This new paradigm is a complete break from traditional data management systems. In this paper we will show a way to define autonomous data as well as some challenges associated with their specificities.
Download

Area 4 - Edge Cloud and Fog Computing

Full Papers
Paper Nr: 23
Title:

Cloud-Native Applications' Workload Placement over the Edge-Cloud Continuum
Authors:

Georgios Kontos, Polyzois Soumplis, Panagiotis Kokkinos and Emmanouel Varvarigos
Abstract: The evolution of virtualization technologies and of distributed computing architectures has inspired the so-called cloud native applications development approach. A cornerstone of this approach is the decomposition of a monolithic application into small and loosely coupled components (i.e., microservices). In this way, application’s performance, flexibility, and robustness can be improved. However, most orchestration algorithms assume generic application workloads that cannot serve efficiently the specific requirements posed by the applications, regarding latency and low communication delays between their dependent microservices. In this work, we develop advanced mechanisms for automating the allocation of computing resources, in order to optimize the service of cloud-native applications in a layered edge-cloud continuum. We initially present the Mixed Integer Linear Programming formulation of the problem. As the execution time can be prohibitively large for real-size problems, we develop a fast heuristic algorithm. To efficiently exploit the performance– execution time trade-off, we employ a novel multi-agent Rollout, the simplest and most reliable among the Reinforcement Learning methods, that leverages the heuristic’s decisions to further optimize the final solution. We evaluate the results through extensive simulations under various inputs that demonstrate the quality of the generated sub-optimal solutions.
Download
Paper Nr: 42
Title:

Towards Optimizing the Edge-to-Cloud Continuum Resource Allocation
Authors:

Igor F. Capeletti, Ariel Goes de Castro, Daniel C. Temp, Paulo S. Severo de Souza, Arthur F. Lorenzon, Fábio D. Rossi and Marcelo C. Luizelli
Abstract: The IT community has witnessed a transition towards the cooperation of two major paradigms, Cloud Computing and Edge Computing, paving the way to a Cloud Continuum, where computation can be performed at the various network levels. While this model widens the provisioning possibilities, choosing the most cost-efficient processing location is not trivial. In addition, network bottlenecks between end users and computing facilities assigned for carrying out processing can undermine application performance. To overcome this challenge, this paper presents a novel algorithm that leverages a path-aware heuristic approach to opportunistically process application requests on compute devices along the network path. Once intermediate hosts process information, requests are sent back to users, alleviating the demand on the network core and minimizing end-to-end application latency. Simulated experiments demonstrate that our approach outperforms baseline routing strategies by a factor of 24x in terms of network saturation reduction without sacrificing application latency.
Download
Paper Nr: 46
Title:

Simulation-Based Estimation of Resource Needs in Fog Robotics Infrastructures
Authors:

Lucien N. Ngale, Eddy Caron, Huaxi Zhang and Mélanie Fontaine
Abstract: Embedded devices are increasingly connected to the Internet to provide new and innovative applications in many areas. These devices (Edge devices or “Things” in IoT) are heterogeneous sensors, cameras and even robots performing sometimes certain tasks locally. Fog Computing (or Fog robotics) optimizes the management of these tasks, offering data management mechanisms (computation and storage) closer to the data source. Nevertheless, many aspects remain closed to Fog computing environments like resource needs estimation in such environments. Indeed, such a topic remains a critical challenge, as it falls under either solving very complex optimization problems or comparing hypothetical scenarios very time consuming and/or expensive for deployment in a real environment. To help on this challenge we built SERFRI, an approach to estimate the resource needs in Fog robotics environments based on simulation. This approach optimizes simultaneously the duration and the Fog resources utilization cost in order to determine the minimum resource requirements compromising both metrics. We validated this approach on an existing robotics use case. This one aims at deploying a human face detection service on streaming images.
Download
Short Papers
Paper Nr: 9
Title:

ARTHUR: Machine Learning Data Acquisition System with Distributed Data Sensors
Authors:

Niels Schneider, Philipp Ruf, Matthias Lermer and Christoph Reich
Abstract: On the way to the smart factory, the manufacturing companies investigate the potential of Machine Learning approaches like visual quality inspection, process optimisation, maintenance prediction and more. In order to be able to assess the influence of Machine Learning based systems on business-relevant key figures, many companies go down the path of test before invest. This paper describes a novel and inexpensive distributed Data Acquisition System, ARTHUR (dAta collectoR sysTem witH distribUted sensoRs), to enable the collection of data for AI-based projects for research, education and the industry. ARTHUR is arbitrarily expandable and has so far been used in the field of data acquisition on machine tools. Typical measured values are Acoustic Emission values, force plate X-Y-Z force values, simple SPS signals, OPC-UA machine parameters, etc. which were recorded by a wide variety of sensors. The ARTHUR system consists of a master node, multiple measurement worker nodes, a local streaming system and a gateway that stores the data to the cloud. The authors describe the hardware and software of this system and discuss its advantages and disadvantages.
Download
Paper Nr: 30
Title:

Self-Healing Misconfiguration of Cloud-Based IoT Systems Using Markov Decision Processes
Authors:

Areeg Samir and Håvard Dagenborg
Abstract: Misconfiguration of IoT devices and backend containerized-cluster systems can expose vulnerable areas at the network level, potentially allowing attackers to penetrate the network and disrupt workload and the flow of data between system components. This paper describes a self-healing model based on a Markov decision process that can recover the misconfiguration and its impact on the workload and data flow at the network level. The results show that the proposed controller led to accurate results in performance and reliability.
Download
Paper Nr: 31
Title:

QoS-Aware Task Allocation and Scheduling in Three-Tier Cloud-Fog-IoT Architecture Using Double Auction
Authors:

Nikita Joshi and Sanjay Srivastava
Abstract: Cloud Architecture is available to process these data. Due to the QoS requirements of IoT applications, the perishable nature of fog cloud resources, and the competition among users and service providers, task allocation in such an architecture is challenging. In this paper, we propose a competitive bidding (CompBid) strategy and a QoS-based task allocation and scheduling (QoTAS) algorithm using double auctions that aim to maximize user and service provider profit while also satisfying QoS requirements. A remote patient monitoring system is used to compare QoTAS performance to that of two previous studies, DPDA and MADA. In both the truthful and CompBid strategies, QoTAS achieves a higher task allocation ratio and resource utilization than DPDA and MADA. It has 89% more system utility than DPDA and 54% more user utility than MADA. Furthermore, the CompBid strategy increases QoTAS system utility by 25%.
Download
Paper Nr: 34
Title:

Sophos: A Framework for Application Orchestration in the Cloud-to-Edge Continuum
Authors:

Angelo Marchese and Orazio Tomarchio
Abstract: Orchestrating distributed applications on the Cloud-to-Edge continuum is a challenging task, because of the continuously varying node computational resources availability and node-to-node network latency and band- width on Edge infrastructure. Although Kubernetes is today the de-facto standard for container orchestration on Cloud data centers, its orchestration and scheduling strategy is not suitable for the management of time critical applications on Edge environments because it does not take into account current infrastructure state during its scheduling decisions. In this work, we present Sophos, a framework that runs on top of the Kuber- netes platform in order to implement an effective resource and network-aware microservices scheduling and orchestration strategy. In particular, Sophos extends the Kubernetes control plane with a cluster monitor that monitors the current state of the application execution environment, an application configuration controller that continuously tunes the application configuration based on telemetry data and a custom scheduler that de- termines the placement for each microservice based on the run time infrastructure and application states. An evaluation of the proposed framework is presented by comparing it with the default Kubernetes orchestration and scheduling strategy.
Download
Paper Nr: 13
Title:

A Comparison of Synchronous and Asynchronous Distributed Particle Swarm Optimization for Edge Computing
Authors:

Riccardo Busetti, Nabil El Ioini, Hamid R. Barzegar and Claus Pahl
Abstract: Edge computing needs to deal with concerns such as load balancing, resource provisioning, and workload placement as optimization problems. Particle Swarm Optimization (PSO) is a nature-inspired stochastic opti- mization approach that aims at iteratively improving a solution of a problem over a given objective. Utilising PSO in a distributed edge setting would allow the transfer of resource-intensive computational tasks from a central cloud to the edge, this providing a more efficient use of existing resources. However, there are chal- lenges to meet performance and fault tolerance targets caused by the resource-constrained edge environment with a higher probability of faults. We introduce here distributed synchronous and asynchronous variants of the PSO algorithm. These two forms specifically target the performance and fault tolerance requirements in an edge network. The PSO algorithms distribute the load across multiple nodes in order to effectively realize coarse-grained parallelism, resulting in a significant performance increase.
Download

Area 5 - Service Modelling and Analytics

Short Papers
Paper Nr: 8
Title:

An Experimental Evaluation of Relations Between Architectural and Runtime Metrics in Microservices Systems
Authors:

Niels Knoll and Robin Lichtenthäler
Abstract: The decisions made about the architecture of a microservices system at design time influence the runtime behavior of the resulting system and can be hard to change later. But predicting or evaluating how excatly architecture decisions impact runtime behavior is difficult and in practice mostly based on previous experience. Architectural metrics that are measurable at design time and have a traceable impact on runtime metrics could support architectural decision making to improve quality and prevent costly redevelopments. To investigate traceable relations between architectural metrics and runtime metrics, this paper presents a model-driven generation system for microservice architectures. The system can be used to benchmark different architecture alternatives of a Java-based application without manually changing application code. Using this system, we performed experiments to examine relations between architectural metrics and throughput as a runtime metric.
Download

Area 6 - Services Science

Full Papers
Paper Nr: 4
Title:

Towards Security-Aware Microservices: On Extracting Endpoint Data Access Operations to Determine Access Rights
Authors:

Amr S. Abdelfattah, Micah Schiewe, Jacob Curtis, Tomas Cerny and Eunjee Song
Abstract: Security policies are typically defined centrally for a particular system. However, the current mainstream architecture - microservices - introduces decentralization with self-contained interacting parts. This brings better evolution autonomy to individual microservices but introduces new challenges with consistency. The most basic security perspective is the setting of access rights; we typically enforce access rights at system endpoints. Given the self-contained and decentralized microservice nature, each microservice has to implement these policies individually. Considering that different development teams are involved in microservice development, likely the access rights are not consistently implemented across the system. Moreover, as the system evolves, it can quickly become cumbersome to identify a holistic view of the full set of access rights applied in the system. Various issues can emerge from inconsistent settings and potentially lead to security vulnerabilities and unintended bugs, such as incorrectly granting write or read access to system data. This paper presents an approach aiding a human-centered access right analysis of system endpoints in microservices. It identifies the system data that a particular endpoint accesses throughout its call paths and determines which operations are performed on these data across the call paths. In addition, it takes into account inter-service communication across microservices, which brings a great and novel instrument to practitioners who would otherwise need to perform a thorough code review of self-contained codebases to extract such information from the system. The presented approach has broad potential related to security analysis, further detailed in the paper.
Download
Paper Nr: 18
Title:

Semi-Automated Smell Resolution in Kubernetes-Deployed Microservices
Authors:

Jacopo Soldani, Marco Marinò and Antonio Brogi
Abstract: Microservices are getting commonplace, as their design principles enable obtaining cloud-native applications. Ensuring that applications adheres to microservices’ design principles is hence crucial, and this includes resolving architectural smells possibly denoting violations of such principles. To this end, in this paper we propose a semi-automated methodology for resolving architectural smells in microservices applications deployed with Kubernetes. Our methodology indeed automatically detects architectural smells by analyzing the Kubernetes manifest files specifying an application’s deployment, and it can also generate the refactoring templates for resolving such smells. We also introduce KubeFreshener, an open-source prototype of our methodology, which we use to assess it in practice based on a controlled experiment and a case study.
Download
Short Papers
Paper Nr: 2
Title:

Supporting Disconnected Operation of Stateful Services Using an Envoy Enabled Dynamic Microservices Approach
Authors:

Tim Farnham
Abstract: Dynamic microservice and service mesh approaches provide many benefits and flexibility for deploying services and setting policies for access control, throttling, load balancing, retry, circuit breaker or shadow mirror configurations. This paper examines extending this to support continuous operation of stateful microservices in hybrid cloud / edge deployment, without loss of data, by permitting disconnected operation and resynchronisation. These are important considerations for critical applications which must continue to operate even during prolonged cloud disconnection and node or client failure. Such service requirements are typical of retail and other scenarios in which services must run continuously, while maintaining a consistent state between cloud and edge service instances. The approach taken and evaluated in this paper exploits a lightweight Envoy proxy within Choreo connect microgateways and Consul service mesh sidecars. Envoy proxies are able to efficiently perform shadow mirroring of requests and support graceful failover, but requires additional functionality to support resynchronisation and recovery from failure that are examined in this paper.
Download
Paper Nr: 12
Title:

Filling The Gaps in Microservice Frontend Communication: Case for New Frontend Patterns
Authors:

Amr S. Abdelfattah and Tomas Cerny
Abstract: Microservices architecture has exploded in popularity; many organizations use this architectural style to avoid the limitations of large and monolithic backends. Most systems require multiple frontend clients, such that each frontend client expects tailored responses from a backend service. However, there are no best practices for their integration and communication with microservice backends. Backend for Frontends (BFF) is one of the most used patterns for gluing the frontend with the microservices layer. It keeps the frontend layer decoupled from the microservices complications; nevertheless, it is tightly coupled with the frontend layer. Therefore, it introduces barriers in the development process, besides adding risks for business inconsistency. In addition, it negatively impacts the consumed overall data size and time over requests. This risk is boosted by the evolution of the micro-frontend architectural style that encourages the decomposition approach for the frontend components. This paper proposes an alternative pattern that addresses current gaps introduced by the BFF patterns. It supports cloud-native system components to provide the required customization to frontends, along with increasing the frontend awareness to share more responsibilities in the architecture. The new pattern facilitates customizability for client types when interacting with the microservices business layer.
Download
Paper Nr: 24
Title:

Microservices Architecture Language for Describing Service View
Authors:

Luka Lelovic, Michael Mathews, Amr S. Abdelfattah and Tomas Cerny
Abstract: Microservices Architecture is a growing trend in recent years that has been promoted due to a number of researched advantages. However, as microservice systems grow and evolve, they can become complex and hard to understand. In order to face this problem, techniques to reconstruct, describe and visualize these systems are proposed. Despite this, there are currently no architectural languages actively maintained, adopted, and promoted as the intermediate between the system reconstruction and its corresponding viewpoints. This paper proposes a YAML-based architectural language acting as the intermediate representation for microservice architecture, specifically in the service view architectural perspective. This paper outlines the new language, its basis, example descriptions, and possible architectural visualizations of the descriptions. It also details how it compares to other existing architectural languages in the microservice domain.
Download
Paper Nr: 43
Title:

QuantME4VQA: Modeling and Executing Variational Quantum Algorithms Using Workflows
Authors:

Martin Beisel, Johanna Barzen, Marvin Bechtold, Frank Leymann, Felix Truger and Benjamin Weder
Abstract: The execution of quantum algorithms typically requires classical pre- and post-processing, making quantum applications inherently hybrid. Classical resources are of particular importance for so-called variational quantum algorithms, as they leverage classical computational power to overcome the limitations imposed by today’s noisy quantum devices. However, the additional complex tasks required by these algorithms, complicate the challenge of integrating quantum circuit executions with classical programs. To overcome this challenge, we leverage the advantages and versatility of workflows, and introduce a workflow modeling extension for orchestrating variational quantum algorithms. The modeling extension comprises new task types, data objects, and a comprehensible graphical notation. Furthermore, we ensure interoperability and portability by providing a method for transforming all new modeling constructs to native modeling constructs, and showcase the practical feasibility of our modeling extension by presenting a system architecture, prototype, and case study.
Download
Paper Nr: 3
Title:

Optimal Static Bidding Strategy for Running Jobs with Hard Deadline Constraints on Spot Instances
Authors:

Kai-Siang Wang, Cheng-Han Hsieh and Jerry Chou
Abstract: Spot-instances(SI) is an auction-based pricing scheme used by cloud providers. It allows users to place bids for spare computing instances and rent them at a substantially lower price compared to the fixed on-demand price. This inexpensive computational power is at the cost of availability, because a spot instance can be revoked whenever the spot market price exceeds the bid. Therefore, SI has become an attractive option for applications without requiring real-time availability constraints, such as the batch jobs in different application domains, including big data analytics, scientific computing, and deep learning. For batch jobs, service interruptions and execution delays can be tolerated as long as their service quality is gauged by an execution deadline. Hence, this paper aims to develop a static bidding strategy for minimizing the monetary cost of a batch job with hard deadline constraints. We formulate the problem as a Markov chain process and use Dynamic Programming to find the optimal bid in polynomial time. Experiments conducted on real workloads from Amazon Spot Instance historical prices show that our proposed strategy successfully outperformed two state-of-art dynamic bidding strategies (Amazing, DBA), and several deadline agnostic static bidding strategies with lower cost.
Download
Paper Nr: 7
Title:

Data Driven Meta-Heuristic-Assisted Approach for Placement of Standard IT Enterprise Systems in Hybrid-Cloud
Authors:

Andrey Kharitonov, Abdulrahman Nahhas, Hendrik Müller and Klaus Turowski
Abstract: We address the problem of hybrid-cloud placement selection for commercial off-the-shelf IT enterprise applications with the sizing done based on workload profiles collected from real-world production systems. The proposed approach leverages techniques based on evolutionary meta-heuristics with a multi-criteria weighted sum objective function. A placement decision is made between an on-premises data center and a public cloud, using real pricing information for virtual machines, storage, and networking published by the public cloud vendor via automation APIs and on-premises cost estimation as a share of expense per service. Additional objectives, such as expertise and non-functional requirements, are encoded in a numerical form for the objective function. The evaluation is performed as single and multi-objective optimization by employing genetic algorithm, and non-dominated-sorting genetic-algorithm-III on the case study of an SAP landscape hybrid-cloud placement on a selected public cloud with real workload data collected during day-to-day business operations, indicating the viability of the approach.
Download
Paper Nr: 16
Title:

Edge Anomaly Detection Framework for AIOps in Cloud and IoT
Authors:

Pieter Moens, Bavo Andriessen, Merlijn Sebrechts, Bruno Volckaert and Sofie Van Hoecke
Abstract: Artificial Intelligence for IT Operations (AIOps) addresses the rising complexity of cloud computing and Internet of Things by assisting DevOps engineers to monitor and maintain applications. Machine Learning is an essential part of AIOps, enabling it to perform Anomaly Detection and Root Cause Analysis. These techniques are often executed in centralized components, however, which requires transferring vast amounts of data to a central location. This increase in network traffic causes strain on the network and results in higher latency. This paper leverages edge computing to address this issue by deploying ML models closer to the monitored services, reducing the network overhead. This paper investigates two architectural approaches: a sidecar architecture and a federated architecture, and highlights their advantages and shortcomings in different scenarios. Taking this into account, it proposes a framework that orchestrates the deployment and management of distributed edge ML models. Additionally, the paper introduces a Python library to assist data scientists during the development of AIOps techniques and concludes with a thorough evaluation of the resulting framework towards resource consumption and scalability. The results indicate up to 98.3% reduction in network usage depending on the configuration used while maintaining a minimal increase in resource usage at the edge.
Download

Area 7 - Cloud Computing Platforms and Applications

Full Papers
Paper Nr: 28
Title:

FaaS Benchmarking over Orama Framework's Distributed Architecture
Authors:

Leonardo Rebouças de Carvalho, Bruno Kamienski and Aleteia Araujo
Abstract: As the adoption of Function-as-a-Service-oriented solutions grows, interest in tools that enable the execution of benchmarks on these environments increases. The Orama framework stands out in this context by offering a highly configurable and scalable solution to aid in provisioning, running benchmarks and comparative and statistical analysis of results. In this work, a distributed architecture for the Orama framework is presented, through which it is possible to run benchmarks with high levels of concurrency, as well as with bursts of geographically dispersed requests, just as in real environments. The results of the experiment showed that the proposed architecture was able to divide the loads between the distributed workers and able to consolidate properly in the return of the results. In addition, it was possible to observe specific characteristics of the providers involved in the experiment, such as the excellent performance of Alibaba Function, whose average execution time was the lowest of the tests and free of failures. Google Cloud Function and AWS Lambda recorded intermediate results for average execution time and recorded failures. Finally, Azure Function had the worst results in average execution time and cold start.
Download
Short Papers
Paper Nr: 6
Title:

An Evaluation Method and Comparison of Modern Cluster-Based Highly Available Database Solutions
Authors:

Raju Shrestha and Tahzeena Tandel
Abstract: High availability in cloud computing is a top concern which refers to keeping a service operational and available for the maximum amount of time without downtime. In any given application and service, the high availability of the database plays a critical role in the application’s high availability as a whole. Modern cluster-based multi-master technologies provide high availability database solutions through synchronous replication. Different database management systems offer different technologies and solutions for high availability. This paper proposes a comprehensive method for the evaluation of high availability database solutions. As a comparative case study, two modern high availability database solutions which are open-source and available for free use, namely the Percona XtraDB Cluster and the MySQL NDB Cluster are used. Benchmark tests with standard CRUD database operations and analysis of the test results show that no single solution is superior to the other in all scenarios and needs. Therefore, one should choose an appropriate solution wisely depending on the needs for an application or service. The proposed evaluation method would be useful to get insights into which solution is effective for a given application and hence it can be used in making an informed choice among different solutions at hand.
Download
Paper Nr: 11
Title:

How to Select Quantum Compilers and Quantum Computers Before Compilation
Authors:

Marie Salm, Johanna Barzen, Frank Leymann and Philipp Wundrack
Abstract: Quantum computers might solve specific problems faster than classical computers in the future. But their actual qubit numbers are small, and the error rates are high. However, quantum computers are already used in various areas and a steadily increasing number is made available by cloud providers. To execute a quantum circuit, it is mapped to the quantum computer’s hardware. The resulting compiled circuit strongly influences the precision of the execution in terms of occurring errors caused by used qubits and quantum gates. Selecting an optimal one is, therefore, essential. SDKs are used to implement circuits and differ in supported cloud providers and programming languages. These differences complicate a change to other backends. In previous work, we developed an automated framework to translate a given circuit and compile it on available quantum computers using multiple compilers. The compilation results can be prioritized and executed. Nevertheless, the translation and compilation with all compilers and quantum computers is resource-intensive and does not scale well with further backends in the future. We, therefore, present an extension to automatically select suitable compiler and quantum computer combinations based on the user’s needs, e.g., for short waiting times and precise results based on past executions. To demonstrate and validate our approach, we show a prototype and case study.
Download
Paper Nr: 37
Title:

A Security Evaluation Framework for Software-Defined Network Architectures in Data Center Environments
Authors:

Igor Ivkić, Dominik Thiede, Nicholas Race, Matthew Broadbent and Antonios Gouglidis
Abstract: Data Center (DC) network requirements. Virtualisation is one of the key drivers of that transformation and enables a massive deployment of computing resources, which exhausts server capacity limits. Furthermore, the increased network endpoints need to be handled dynamically and centrally to facilitate cloud computing functionalities. Traditional DCs barely satisfy those demands because of their inherent limitations based on the network topology. Software-Defined Networks (SDN) promise to meet the increasing network requirements for cloud applications by decoupling control functionalities from data forwarding. Although SDN solutions add more flexibility to DC networks, they also pose new vulnerabilities with a high impact due to the centralised architecture. In this paper we propose an evaluation framework for assessing the security level of SDN architectures in four different stages. Furthermore, we show in an experimental study, how the framework can be used for mapping SDN threats with associated vulnerabilities and necessary mitigations in conjunction with risk and impact classification. The proposed framework helps administrators to evaluate the network security level, to apply countermeasures for identified SDN threats, and to meet the networks security requirements.
Download
Paper Nr: 36
Title:

AI Marketplace: Serving Environment for AI Solutions Using Kubernetes
Authors:

Marc A. Riedlinger, Ruslan Bernijazov and Fabian Hanke
Abstract: Recent advances in the field of artificial intelligence (AI) provide numerous potentials for industrial compa- nies. However, the adoption of AI in practice is still left behind. One of the main reasons is a lack of knowledge about possible AI application areas by industry experts. The AI Marketplace addresses this problem by pro- viding a platform for the cooperation between industry experts and AI developers. An essential function of this platform is a serving environment that allows AI developers to present their solution to industry experts. The solutions are packaged in a uniform way and made accessible to all platform members via the serving environment. In this paper, we present the conceptual design of this environment, its implementation using Amazon Web Services, and illustrate its application on two exemplary use cases.
Download
Paper Nr: 45
Title:

Accelerating Deep Learning Model Training on Cloud Tensor Processing Unit
Authors:

Cristiano A. Künas, Edson L. Padoin and Philippe A. Navaux
Abstract: Deep learning techniques have grown rapidly in recent years due to their success in image classification, speech recognition, and natural language understanding. These techniques have the potential to solve complex problems and are being applied in various fields, such as agriculture, medicine, and administration. However, training large and complex models requires high-performance computational platforms, making accelerator hardware an essential tool and driving up its cost. An alternative solution is to use cloud computing, where users only pay for usage and have access to a wide range of computing resources and services. In this paper, we adapt a Diabetic Retinopathy neural network model for TPU-based training in the cloud and observe promising results, including reduced training time without code optimization. This demonstrates the potential of cloud computing in reducing the burden on local systems that are often overwhelmed by multiple running applications. This allows for training larger and more advanced models at a lower cost than local computational centers.
Download
Paper Nr: 47
Title:

Empirical Analysis of Federated Learning Algorithms: A Federated Research Infrastructure Use Case
Authors:

Harshit Gupta, Abhishek Verma, O. P. Vyas, Marco Garofalo, Giuseppe Tricomi, Francesco Longo, Giovanni Merlino and Antonio Puliafito
Abstract: Research Infrastructures provide resources and services for communities of researchers at large to conduct their experiments and foster innovation. Moreover, these can be used beyond research, e.g., for education or public service. The SLICES consortium is chartered to provide a fully programmable, distributed, virtualized, remotely accessible, European-wide, federated research infrastructure, providing advanced computing, storage, and networking capabilities, including interconnection by dedicated high-speed links. It will support large-scale, experimental research across various scientific domains. Data processing, in general, and especially Machine Learning, are of great interest to the potential audience of SLICES. According to these premises, this work aims to exploit such a peculiar Research Infrastructure and its Cloud-oriented development and deployment facilities to investigate Federated Learning (FL) approaches; in particular, here we evaluate the performance of two FL aggregation algorithms, i.e., FedAvg and FedProx, in settings, characterized by system heterogeneity, and statistical heterogeneity, that represent plausible, and possibly common, scenarios in forthcoming facilities, such as those mentioned above, community-oriented, shared Research Infrastructures. We have observed that the FedProx algorithm outperforms the FedAvg algorithm in such settings.
Download

Area 8 - Cloud Computing Enabling Technology

Full Papers
Paper Nr: 20
Title:

YASF: A Vendor-Agnostic Framework for Serverless Computing
Authors:

Mikael Giacomini and Amjad Ullah
Abstract: The serverless execution model allows application developers to deploy their software using tiny functions with zero administration, no handling of resource provisioning, monitoring and scaling. Due to such advantages, the serverless model emerged as a new promising paradigm, where pay as you go offerings can be found by all public cloud providers. However, these offerings encourage vendor lock-in. This paper aims to address the vendor lock-in issue using a novel framework that combines the strength of an agnostic infrastructure configuration based on the Constructs Programming Model, and the creation of abstraction layers that supports function logic by handling provider-specific integration. The proposed framework abstracts away the specificities and complications of the various underlying serverless platforms and the application developers are only required to provide their specific functions. The evaluation consists of the deployment of a benchmark application across different cloud providers to demonstrate the ease and flexibility of the framework.
Download
Paper Nr: 39
Title:

Enabling Quantum Key Distribution on a Multi-Cloud Environment to Secure Distributed Data Lakes
Authors:

Jose L. Lo Huang and Vincent C. Emeakaroha
Abstract: Each day more and more data is produced by different sources including humans and machines, and they are stored mainly in Cloud Service Providers (CSP) in huge data storages known as big data or data lakes. Many situations warrant users to spread their data in a distributed way between the different CSP. When this happens, they have to relay the inter-cloud communication security to each cloud vendor. This can cause data leakage in the transmission channel thereby compromise information security. Quantum computing has shown some promises to address this issue. One well known algorithm in quantum cryptography is the Quantum Key Distribution (QKD) protocol. This enables the sender and receiver of a message to know when a third party eavesdropped any data from the insecure quantum channel. There are studies integrating this QKD protocol with cloud storage and data transmission inside one CSP. However, there is no research that studies the data lake security concern for distributed multi-cloud communications taking advantage of quantum mechanisms. This research proposal aims to address this gap in distributed data lake security by using the QKD protocol in the multi-cloud distributed data transmission. The achieved results show over 91% detection of eavesdropping cases and over 99% correct authorisation detection in multi-cloud environments.
Download
Short Papers
Paper Nr: 22
Title:

Customisable Fault and Performance Monitoring Across Multiple Clouds
Authors:

Giuseppe Bisicchia, Stefano Forti, Alberto Colla and Antonio Brogi
Abstract: Monitoring the proper functioning and performance of an infrastructure spanning multiple Cloud datacentres is challenging. It requires continuously aggregating monitored data across multiple source machines and processing them so to obtain useful alerts and insights. In this article, we propose a simple open-source prototype tool to perform highly customisable fault and performance monitoring across multiple Clouds. Differently from commercial tools, our prototype is simpler to deploy and it can be configured through a declarative approach, by simply specifying data monitoring tasks and aggregation policies. We illustrate such peculiarities over a use case relying on three datacentres under the Italian Research and Education Network Consortium.
Download
Paper Nr: 29
Title:

A Recent Publications Survey on Reinforcement Learning for Selecting Parameters of Meta-Heuristic and Machine Learning Algorithms
Authors:

Maria Chernigovskaya, Andrey Kharitonov and Klaus Turowski
Abstract: Nowadays, meta-heuristic and machine learning algorithms are often used for a variety of tasks in cloud computing operations. The choice of hyper-parameter values has a direct impact on the performance of these algorithms, making Hyper-Parameter Optimization (HPO) an important research field for facilitating the widespread application of machine learning and meta-heuristics for problem-solving. Manual parameter- ization of these algorithms is an inefficient method, which motivates researchers to look for a new and more efficient approach to tackle this challenge. One such innovative approach is Deep Reinforcement Learning (DRL), which has recently demonstrated a lot of potential in solving complex problems. In this work, we aim to explore this topic more thoroughly and shed light on the application of DRL-based techniques in HPO, specifically for Machine Learning and Heuristics/Meta-heuristics-based algorithms. We approach the problem by conducting a systematic literature review of the recently published literature and summarizing the results of the analysis. Based on the conducted literature review, within the selected sources, we identified 14 relevant publications and a clear research gap in the cloud-specific use case for HPO via DRL.
Download