RECAP Work Package 8 (WP8) is focused on solving application placement and infrastructure optimisation challenges, for initial resource allocation and capacity planning. This deliverable outlines the design of initial models, abstractions, and algorithms that have been developed towards achieving this goal. This functionality has been encapsulated as part of the 'Placement & Infrastructure Optimiser' architecture component, an integral element of the integrated RECAP architecture.
For each RECAP use case, the deliverable presents optimisation objectives, and explains the trade-offs to be considered between them. It further represents the use cases as service-to-resource mappings, explaining the terminology and the constraints considered. To balance edge resource allocations providing low latency but constrained in capacity, with cloud resource allocations providing high capacity at the cost of a higher latency, a multi-objective model is formulated. The model quantifies these competing objectives using utility functions, as common unifying mathematical metrics, and applies the Multi-Attribute Utility Theory to support trade-offs and prioritizations between objectives.
These formulations support optimally matching the requirements of different service components to the capabilities of the infrastructure resources, thereby instantiating the components with the allocated capacity, and finally, adhering to Service Level Agreements (SLAs) negotiated between the infrastructure provider and the service provider.
The design and initial validation of a heuristic – based on the defined model and evolutionary algorithms – is presented to determine optimal application placements for RECAP use cases. While the algorithm is designed to address RECAP use cases, it can also be applied to other applications and application domains facing similar application placement and infrastructure optimization challenges.
Initial experimentation and analysis for the BT vCDN use case is presented, with the deliverable also documenting progress on the Tieto use case. These use cases were prioritized as they both have infrastructure and application KPIs providing opportunities for trade-off between objectives.
Results show initial characterization of the BT vCDN use case supporting translation of application behaviour into resource and capacity requirements. The integration and trade-off between four utility functions is also presented to capture the impact of the application, express correlations between parameters including utilisation, cost and latency, and support infrastructure capacity planning.
To download this deliverable, please fill in the form.