Stefano Sebastio

A computational field framework for collaborative task execution in volunteer clouds

The increasing diffusion of cloud technologies offers new opportunities for distributed and collaborative computing. Volunteer clouds are a prominent example, where participants join and leave the platform and collaborate by sharing computational resources. The high complexity, dynamism and unpredictability of such scenarios call for decentralized self-* approaches. We present in this paper a framework for the design and evaluation of self-adaptive collaborative task execution strategies in volunteer clouds. As a byproduct, we propose a novel strategy based on the Ant Colony Optimization paradigm, that we validate through simulation-based statistical analysis over Google cluster data.

A Cooperative Approach for Distributed Task Execution in Autonomic Clouds

Virtualization and distributed computing are two key pillars that guarantee scalability of applications deployed in the Cloud. In Autonomous Cooperative Cloud-based Platforms, autonomous computing nodes cooperate to offer a PaaS Cloud for the deployment of user applications. Each node must allocate the necessary resources for applications to be executed with certain QoS guarantees. If the QoS of an application cannot be guaranteed a node has mainly two options: to allocate more resources (if it is possible) or to rely on the collaboration of other nodes. Making a decision is not trivial since it involves many factors (e.g. the cost of setting up virtual machines, migrating applications, discovering collaborators). In this paper we present a model of such scenarios and experimental results validating the convenience of cooperative strategies over selfish ones, where nodes do not help each other. We describe the architecture of the platform of autonomous clouds and the main features of the model, which has been implemented and evaluated in the DEUS discrete-event simulator. From the experimental evaluation, based on workload data from the Google Cloud Backend, we can conclude that (modulo our assumptions and simplifications) the performance of a volunteer cloud can be compared to that of a Google Cluster.

Distributed statistical analysis of complex systems modeled through a chemical metaphor

The chemical-inspired programming approach is an emerging paradigm for defining the behavior of densely distributed and context-aware devices (e.g., in ecosystems of displays tailored to crowd steering, or to obtain profile-based coordinated visualization). Typically, the evolution of such systems cannot be easily predicted, thus making of paramount importance the availability of techniques and tools supporting prior-to-deployment analysis. Exact analysis techniques do not scale well when the complexity of systems grows: as a consequence, approximated techniques based on simulation assumed a relevant role. This work presents a new simulation-based distributed analysis tool addressing the statistical analysis of such a kind of systems. The tool has been obtained by chaining two existing tools: MultiVeSta and Alchemist. The former is a recently proposed lightweight tool which allows to enrich existing discrete event simulators with automated and distributed statistical analysis capabilities, while the latter is an efficient simulator for chemical-inspired computational systems. The tool is validated against a crowd steering scenario, and insights on the performance are provided by discussing how the analysis tasks scale on a multi-core architecture.

Fast computation of bounds for two-terminal network reliability

In this paper, an algorithm for the fast computation of network reliability bounds is proposed. The evaluation of the network reliability is an intractable problem for very large networks, and hence approximate solutions based on reliability bounds have assumed importance. The proposed bounds computation algorithm is based on an efficient BDD representation of the reliability graph model and a novel search technique to find important minpaths/mincuts to quickly reduce the gap between the reliability upper and lower bounds. Furthermore, our algorithm allows the control of the gap between the two bounds by controlling the overall execution time. Therefore, a trade-off between prediction accuracy and computational resources can be easily made in our approach. The numerical results are presented for large real example reliability graphs to show the efficacy of our approach.

AVOCLOUDY: a simulator of volunteer clouds

The increasing demand of computational and storage resources is shifting users toward the adoption of cloud technologies. Cloud computing is based on the vision of computing as utility, where users no more need to buy machines but simply access remote resources made available on‐demand by cloud providers. The relationship between users and providers is defined by a service‐level agreement, where the non‐fulfillment of its terms is regulated by the associated penalty fees. Therefore, it is important that the providers adopt proper monitoring and managing strategies. Despite their reduced application, intelligent agents constitute a feasible technology to add autonomic features to cloud operations. Furthermore, the volunteer computing paradigm—one of the Information and Communications Technology (ICT) trends of the last decade—can be pulled alongside traditional cloud approaches, with the purpose to ‘green’ them. Indeed, the combination of data center and volunteer resources, managed by agents, allows one to obtain a more robust and scalable cloud computing platform. The increased challenges in designing such a complex system can benefit from a simulation‐based approach, to test autonomic management solutions before their deployment in the production environment. However, currently available simulators of cloud platforms are not suitable to model and analyze such heterogeneous, large‐scale, and highly dynamic systems. We propose the AVOCLOUDY simulator to fill this gap. This paper presents the internal architecture of the simulator, provides implementation details, summarizes several notable applications, and provides experimental results that measure the simulator performance and its accuracy. The latter experiments are based on real‐world worldwide distributed computations on top of the PlanetLab platform. Copyright

Reputation-Based Cooperation in the Clouds

The popularity of the cloud computing paradigm is opening new opportunities for collaborative computing. In this paper we tackle a fundamental problem in open-ended cloud-based distributed computing platforms, i.e., the quest for potential collaborators. We assume that cloud participants are willing to share their computational resources for shared distributed computing problems, but they are not willing to disclose the details of their resources. Lacking such information, we advocate to rely on reputation scores obtained by evaluating the interactions among participants. More specifically, we propose a methodology to assess, at design time, the impact of different (reputation-based) collaborator selection strategies on the system performance. The evaluation is performed through statistical analysis on a volunteer cloud simulator.

A Holistic Approach for Collaborative Workload Execution in Volunteer Clouds

The demand for provisioning, using, and maintaining distributed computational resources is growing hand in hand with the quest for ubiquitous services. Centralized infrastructures such as cloud computing systems provide suitable solutions for many applications, but their scalability could be limited in some scenarios, such as in the case of latency-dependent applications. The volunteer cloud paradigm aims at overcoming this limitation by encouraging clients to offer their own spare, perhaps unused, computational resources. Volunteer clouds are thus complex, large-scale, dynamic systems that demand for self-adaptive capabilities to offer effective services, as well as modeling and analysis techniques to predict their behavior. In this article, we propose a novel holistic approach for volunteer clouds supporting collaborative task execution services able to improve the quality of service of compute-intensive workloads. We instantiate our approach by extending a recently proposed ant colony optimization algorithm for distributed task execution with a workload-based partitioning of the overlay network of the volunteer cloud. Finally, we evaluate our approach using simulation-based statistical analysis techniques on a workload benchmark provided by Google. Our results show that the proposed approach outperforms some traditional distributed task scheduling algorithms in the presence of compute-intensive workloads.

A Workload-Based Approach to Partition the Volunteer Cloud

The growing demand of computational resources has shifted users towards the adoption of cloud computing technologies. Cloud allows users to transparently access to remote computing capabilities as an utility. The volunteer computing paradigm, another ICT trend of the last years, can be considered a companion force to enhance the cloud in fulfilling specific domain requirements, such as computational intensive requests. Combining the spared resources provided by volunteer nodes with few data centers is possible to obtain a robust and scalable cloud platform. The price for such benefits relies in increased challenges to design and manage a dynamic complex system composed by heterogeneous nodes. Task execution requests submitted in the volunteer cloud are usually associated with Quality of Service requirements e.g., Specified through an execution deadline. In this paper, we present a preliminary evaluation of a cloud partitioning approach to distribute task execution requests in volunteer cloud, that has been validated through a simulation-based statistical analysis using the Google workload data trace.

A Mean Field Model of Coupled Cascades in Flow Networks

We introduce an analytical model of cascading behavior of interdependent networks under stressing conditions and find evidence of abrupt breakdown phenomena. Our results indicate that coupling several infrastructures can diminish the impact of small cascades at the cost of increasing system wide ones. As a consequence, the enhancement of the systemic risk failures with increasing network size, represents an effect to be accounted while planning projects aiming to integrate national networks into “super-networks”.

Adopting the cloud to manage the electricity grid

Public utilities, such as electricity, telecommunication, natural gas, water or sewage, constitute services whose proper functioning is of paramount importance for the whole society. Among these critical infrastructures, power grid and telecommunication network are, probably, the most critical ones. Their interdependency could exacerbate the consequence of a failure since the telecommunication network devices are powered by the mains, while the nationwide power grid is managed through a SCADA system that relies on the public telecommunication network. Cloud technology could reduce costs and improve performance but its adoption should be cautiously evaluated relatively to a critical infrastructure. We qualitatively discuss the cloud adoption in such a kind of infrastructure. Then, a topological analysis, concerning network reliability and robustness, is performed focusing on the SCADA system of the Italian power transmission network. Results show that a certain cloud configuration could be beneficial for the power transmission network.