D7.4. Final Quality of Service Metrics and Models

A fundamental challenge posed by cloud-edge computing capacity provisioning is the fulfilment of Quality of Service (QoS) requirements given a collection of diverse types of services deployed as chains of components across different regions and physical infrastructures. The fulfilment of the end-to-end QoS requirements depends on the QoS of these individual components, and so, the definition of metrics and QoS models is a vital mechanism to enable the continued availability of the services and to improve hardware utilization.

Work package WP7 Large Scale Simulation Framework provides a simulation framework for analysis and prediction of trade-offs involving cost, energy, resource allocation, resource utilization and performance to evaluate different placement and scheduling options close to real time. In particular, Task T7.3 Quality of Service Specification and Modelling provides formal definitions of metrics and specifications of quality-of-service models for applications in next generation cloud systems.

In deliverable D7.3 Initial Quality of Service Metrics and Models, an initial collection of definitions of QoS metrics was proposed, and a first version of the QoS models were formulated. D7.3 also showed how these metrics and models can be applied for workload, application and infrastructure models. In the present deliverable D7.4 Final Quality of Service Metrics and Models the QoS metrics have been updated to reflect the latest requirements gathered from the use cases, and the QoS models were reformulated and extended to address these requirements and the limitations found in simulations and experiments. Results of the new theoretical developments are shown in application for the use cases A (Infrastructure and Network Management), B (Big Data Analytics Engine) and C (Edge/Fog Computing for Smart Cities). Use Cases D1 and D2 were already fully covered in Deliverable D7.3 Initial Quality of Service Metrics and Models and are not reported in this deliverable in order to avoid duplicate contents.