SILICON aims at evaluating, extending and optimizing QoS-Oriented Service Management mechanisms in Large Scale Federated Clouds. The experiment is designed to leverage the outcomes of a leading Internet of Services project (IRMOS) with respect to monitoring, workflow and event management by also incorporating the enhancements in the monitoring mechanism being developed in the framework of the VISION Cloud project. These mechanisms will be evaluated using a computer vision application scenario (developed in the framework of the SCOVIS project), capable to identify, categorize and finally understand visually traced industrial processes and behaviors in real-world industrial environments.
“How scalable can the service be?” is one of the main questions SILICON aims to answer by deploying both cloud and application services in up to one hundred fifty (150) VMs - as well as to on request resources - across six (6) sites. “How does the network affect the service and application behavior?” is another main question SILICON aims to answer by analyzing the network implications both on the cloud services and on the application ones. SILICON will evaluate, extend and optimize the QoS-oriented Service Management mechanisms on a scalable federated Cloud computing environment, with respect to aspects such as: (i) reliability, efficiency and performance of the components in a large scale environment, (ii) feasibility, scalability and compatibility of the proposed components as well as their limitations when deployed in a multi-cloud facility with geographically separated sites in different administrative domains, (iii) network overheads given the hierarchical approach of the service management mechanisms. It is within the main objectives of SILICON to verify and validate the monitoring mechanism of the BonFIRE facility regarding both standard metrics at VM or physical level and user-specified metrics by analyzing aspects such as performance, accuracy, monitoring parameters, network overhead, applicability to different application service components, etc.
Furthermore, the experiment aims at evaluating crosscutting characteristics and functionalities of BonFIRE facility (e.g. multi-site, controlled deployment, heterogeneity, on-demand resource provisioning), enhancing the facility with QoS-oriented mechanisms and providing feedback regarding new requirements and functionalities, areas for improvement and experimentation methodologies.
Added-value of the BonFIRE infrastructure
Experimentation could not be conducted in another (private or public) cloud environment due to the BonFIRE differentiator capabilities, since SILICON will exploit a number of the unique features of the facility, such as: (i) controlled deployment on the multi-site clouds with the event evaluator targeting events for scaling decisions and corresponding deployments under specific timing constraints, (ii) on-demand provision of resources based on scalability rules that impose the deployment of new VMs during application execution, (iii) monitoring of standard metrics at VM or physical level and monitoring of user specified metrics in order to compare and evaluate the SILICON monitoring mechanism with the BonFIRE monitoring framework, (iv) multi-site facility / geographically distributed triggering a feasibility analysis in terms of scalability, dynamicity, performance, network overheads, etc given the “deadline-driven experiment”, (v) controlled Network aiming at understanding the effect of network conditions e.g. best-effort, specific bandwidth limit on the services operation), (vi) heterogeneous resources utilization to evaluate the access and possible limitations to such resources for different configuration parameters.
Impact on the BonFIRE Project
The outcomes of SILICON will contribute to different areas: Infrastructure & Services, Application and Research. SILICON will introduce a solution for QoS-oriented Service Management on geographically distributed sites by developing methods (e.g. adaptable monitoring) for overcoming large-scale facilities bottlenecks and implications on “real-time” applications, proposing approaches for managing the uncertainty introduced due to the heterogeneity of the resources, implementing mechanisms that improve efficiency of experimental infrastructures by advancing their QoS capabilities and by providing outcomes with respect to network implications when deploying “interacting” services. With respect to application services, SILICON will identify how large-scale infrastructure characteristics affect application behavior, propose methods to manage application execution when requesting on demand resources at runtime and provide outcomes with respect to network implications for applications that consist of service components (deployed in interconnected cloud sites). Furthermore, SILICON will develop a methodology for experimentally-driven research following an “increase in realism” approach and publish experiences and best practices developed within the experiment.