Partner:  University of Manchester

The Services Systems Design Group at the University of Manchester is conducting an experiment on BonFIRE, which aims to determine the conditions for achieving resilient and optimal service compositions on a distributed cloud infrastructure for the Future Internet. It will deploy and test two service optimization models, characterized as global optimization and local optimization. The first one, developed in the EC-funded SOA4All project, computes the optimization of a service composition by analysing end-to end interactions between services. Local optimization is driven by the Dynamic Agent-based Ecosystem Model, which computes local optimizations of service compositions by letting the one-to-one interactions between any service provider and any consumer create emergent service chains providing composite services that are resilient to changes.

Added-value of the BonFIRE Infrastructure

The multi-site cloud architecture provided by BonFIRE will help in creating optimal distributed versions of our optimization models and testing their performance. Since both service composition optimization models in TEOS are computationally heavy, and their usability for optimizing compositions involving large number of services is impeded by their computational complexity. Thus, BonFIRE represents an ideal platform for testing the optimizers in various Future Internet configurations e.g. the use of multi-site computational and storage resources. Moreover, the use of BonFIRE provides an opportunity to test the optimizers in real world scenarios involving large number of services e.g. complex supply chain or virtual organization formation that may require multiple computational resources to  manage or balance the computational load.

Impact on the BonFIRE project

The experiments in TEOS aim to test the infrastructure provided by the BonFIRE project. In this respect, the experiments in TEOS are designed to evaluate the various facilities and features of BonFIRE such as the ability of infrastructure to cope with dynamisms of service ecosystem, elasticity, cross-site communication between components and dynamic management of Virtual Machines. Apart from the above, there is opportunity for BonFIRE to learn from the design and use of mechanisms that will be developed and tested in TEOS such as mechanisms for dynamic management of experiments, which is done at experiment runtime via Java software.