The massive proliferation of Internet of Thing (IoT) devices in recent years has created a need for increased computation capability near data sources i.e., at the network edge, for executing low-latency and real-time workloads that process data generated by these sources. Recently, the Cloud-Edge Continuum (CEC) has emerged as a paradigm that can satisfy this need by distributing infrastructural services – hence computation capability – across three distinct architectural layers: the edge layer comprising the ‘data sources’ i.e., resource-constrained heterogeneous devices (e.g., sensors); the cloud layer, comprising resource-rich servers concentrated in data centres; and the intermediary fog layer, comprising heterogeneous servers (e.g., gateways, micro-data centres) in proximity to data sources.
A major challenge in the CEC is service brokerage i.e., the discovery and orchestration of appropriate infrastructural services across the three CEC layers that can optimally satisfy preset quality of service (QoS) requirements, the so-called Service Level Objectives (SLOs), associated with workload execution, such as response time, availability and throughput.
Naturally, before seeking how to satisfy SLOs, they need to be (accurately) described. Nonetheless, describing SLOs in the CEC is a far-from-trivial task that needs to account for the high network and workload variability, hence the unpredictability, near the edge.
Semantic SLAs
In NebulOuS, this challenge is tackled using semantic Service Level Agreements (SLAs). Semantic technologies offer a sophisticated and effective approach to managing the complexities involved in describing QoS aspects in the CEC. They provide interoperability and context awareness through integration of external domain models, as well as the ability to precisely and unambiguously define and interpret composite and contextualized metrics for assessing workload performance. They also enable the definition of accurate compensation schemes tailored to specific contexts and requirements; through semantic reasoning, appropriate compensation actions –ranging from transitioning to different levels of service quality, to ending SLAs with penalties– may be automatically triggered.
Semantic reasoning may also be used to:
- Establish SLA correctness by characterising the inter-relations between the various quality specifications that an SLA comprises i.e., whether they are pairwise mutually contradicting, whether the one subsumes the other, or whether they are orthogonal to each other.
- Assure the quality of the provisioned brokerage service by establishing the compliance of SLAs with higher-order constraints that harness the service-level objectives that these SLAs impose. For instance, a higher-order policy may insist that a workload must stop being consumed if its response time averages more than 2ms over any 30’ period; similarly, it may insist that a workload must stop being consumed if 3 or more violations of its SLOs are observed in any hourly interval. Any SLA covering the consumption of this workload must abide by these higher-order constraints. Higher-order constraints are expressed too ontologically.
Simeon Veloudis
Senior Researcher - South-East European Research Centre (SEERC)
Simeon (Simos) Veloudis is an Associate Professor and the Research Director of the Computer Science Department at CITY College, University of York Europe Campus. He is also a Senior Researcher at the South-East European Research Centre (SEERC). Simos holds a BSc and a PhD degree in Computer Science from the University of Reading. His PhD thesis was on formal (process-algebraic) modelling of safety-critical and real-time systems. Over the last decade, Simos has participated in several EU-funded multimillion-euro R&D projects in such areas as Cloud Computing, Cloud Security, and Responsible Research and Innovation. Simos has worked as an adjunct lecturer in the Computer Science departments of the Technological Institutes of Thessaloniki and Serres. He has published widely in international peer-reviewed journals and conferences, and has participated in numerous Review and Programme committees of journals and conferences. His research interests lie within the realms of Software Engineering, Cloud & Edge Computing, Cybersecurity, and Responsible Research and Innovation.
