Publications

The cloud continuum paradigm promises a transformative approach to integrating the currently fragmented edge and cloud architectures. A comprehensive ecosystem is imperative to truly harness its potential, necessitating the orchestrated integration of diverse features. This article presents five key enablers—dynamic workloads and topology management, data-aware networking, enhanced privacy, zero-emission goal, and flexible resource trading. The synergy of these enablers yields significant advantages, and their practical impact is demonstrated via a compelling use case. As the shift toward computing continuum gains momentum, mastering and integrating these features will define success in the next generation of cloud architectures.

Given the continuously increasing greenhouse gas emissions of information and communication technologies (ICT), there is a pressing need for carbon-aware approaches for computing systems. This study focuses on shifting, which exploits the varying operational carbon intensities across locations and times to move workloads to greener zones. Through a systematic literature review, this research consolidates existing definitions and metrics while identifying significant limitations in current models and practical algorithms. To address these gaps, we propose a novel carbon-aware algorithm for spatio-temporal workload shifting and introduce a unifying definition. Validated through simulations using real-world grid carbon intensity data from across Europe, our ...

Orchestrating services in Edge-Cloud Continua presents significant challenges due to the dynamic nature of resources and user demands. Traditional orchestration approaches struggle to balance performance, cost, and energy efficiency, underscoring the need for advanced automated decision-making. Cognitive orchestration offers a promising solution by continuously observing system state, analyzing performance data, planning adaptations, and executing actions autonomously. This methodology leverages intelligence models to dynamically adapt to evolving requirements and resource availability, aiming to optimize multiple conflicting objectives without human intervention. This paper introduces an initial framework that integrates large language models to facilitate...

The evolving complexity of modern applications, coupled with the rising demand for computation, drives infrastructure providers to distribute their clusters leveraging the computing continuum dynamically. To cope with highly variable workload demands, distributed Kubernetes clusters rely heavily on Cluster Autoscalers (CAs) to provision capacity based on user constraints. However, as providers continue to scale, managing nodes becomes crucial to ensure efficiency, resilience, and adaptability while minimizing operational costs, opening new challenges to CAs, primarily bound to a single cloud provider. This paper outlines six essential properties distinguishing Multi-Provider CAs. Based on a comprehensive review, no existing CA satisfies all these properties. As a...

*Abstract not available via Google Scholar extract. Please refer to the official conference proceedings when published.*

*Abstract not available via Google Scholar extract. Doctoral/Master thesis at Politecnico di Torino.*

The computing continuum combines computational resources and services from edge to cloud, promising enhanced efficiency and resilience with respect to the traditional siloed-based approach. This study presents the REAR (Resource Advertisement and Reservation) protocol, which tackles the complexities of managing resources within this continuum. REAR establishes standardized interfaces to enable interoperability, enhances resource allocation efficiency, and maintains security measures for workload execution. The paper details the protocol’s design, key components, operational workflows, and potential uses, contributing to the optimization of resource use across the computing continuum.

The shift towards agile microservice architecture has enabled significant benefits for IT companies but has also resulted in increased complexity for Cloud orchestration tools. Traditional tools were designed for centralized data centers and are ineffective for locating microservices in geographically-distributed edge-like infrastructures. This paper presents Phare, a novel cloud-native tool designed to natively schedule Kubernetes applications composed of multiple interconnected workloads across federated edge clusters. Our approach addresses the complexity of deciding proper workload placements by analyzing application traffic flows jointly...

Given the ongoing energy crisis, service providers are shifting resource orchestration from a strictly performance-centric viewpoint to one that prioritizes energy-efficient strategies. The edge-cloud continuum further complicates achieving energy efficiency dynamically orchestrating jobs on distributed multi-provider infrastructure. However, state-of-the-art orchestrators lack comprehensive metrics to properly address energy-awareness resource allocations while ensuring the strict multi-provider separation of concerns. To fill this gap, this paper introduces the REAR Energy Profiling model to evaluate the impact of energy-aware workload distribution in the edge-cloud continuum infrastructure. The model proposes a novel methodology allowing peering...

Traditional resource management approaches are characterized by rigid workload assignments, with applications being locked either within the network, edge datacenters, or the cloud. Conversely, the notion of fluid capacity allows dynamic migrations of workloads along the far edge/cloud continuum. The hypothesis investigated in this research is that such fluid capacity allows to dynamically consolidate the components characterizing an edge/cloud continuum into as few locations as possible, bringing high savings in term of overall energy consumption. To test this hypothesis, we developed a mathematical model, which allows to evaluate alternative placement strategies. This article demonstrates how energy efficiency drives the migration of elastic...

Distributed energy generation and the need to accommodate the rapidly evolving changes of energy production and usage represent today huge challenges in modern power grids. Synchrophasors represent one possible cornerstone to improve active control over the grid, to promptly and locally identify changing scenarios or faults. We propose a general, versatile, scalable and software-defined framework that leverages the edge computing paradigm to realize a new synchrophasor application, avoiding to send all the produced data to the cloud, thus limiting latency and bandwidth waste. Performance of this architecture is evaluated, quantifying the number of synchrophasor elements that could be hosted in a typical edge node without sacrificing the strict latency bounds...

Next generation softwarized networks will be dynamically orchestrated over multiple technologies distributed along the edge/cloud continuum and will require the allocation of resources strictly necessary to serve the instantaneous workload required by the service. So-called serverless computing promises to deliver elasticity to arbitrary services, enabling true “pay-as-you-go” models that automatically scale instances with the underlying traffic intensity. This paper proposes a lightweight architecture, RAYGO, to add elasticity traits to generic TCP services running in cloud-native deployments; when no users are active, actual applications are removed, their TCP connections properly hidden, and finally recreated on demand when needed. Preliminary performance...

In recent years, Cloud computing has led a pervasive transformation in application development and execution paradigms. One of these paradigms is the Serverless paradigm or Functions as a Service (FaaS), which completely abstracts the developer from the infrastructural and deployment effort, thus being able to focus only on application logic. Along with these transformations, the use and adoption of Kubernetes as a Cloud orchestration tool, with continuous updates, is increasingly pervasive. This thesis focuses on optimizing Kubernetes for FaaS/Serverless workloads (which typically require fast spin-up of short-lived tasks), analyzing in detail the bottlenecks of the default orchestration workflow and implementing a custom scheduling approach characterized...