The GEYSERS optical testbed: A platform for the integration, validation and demonstration of cloud-based infrastructure services
The recent evolution of cloud services is leading to a new service transformation paradigm to accommodate network infrastructures in a cost-scalable way. In this transformation, the network constitutes the key to efficiently connect users to services and applications. In this paper we describe the deployment, validation and demonstration of the optical integrated testbed for the "GEneralized architecture for dYnamic infrastructure SERviceS" (GEYSERS) project to accommodate such cloud based
... structure Services. The GEYSERS testbed is composed of a set of local physical testbeds allocated in the facilities of the GEYSERS partners. It is built up based on the requirements specification, architecture definition and per--layer development that constitutes the whole GEYSERS ecosystem, and validates the procedures on the GEYSERS prototypes. The testbed includes optical devices (layer 1), switches (layer 2), and IT resources deployed in different local testbeds provided by the project partners and interconnected among them to compose the whole testbed layout. The main goal of the GEYSERS testbed is twofold. On one hand, it aims at providing a validation ground for the architecture, concepts and business models proposed by GEYSERS, sustained by two main paradigms: Infrastructure as a Service (IaaS) and the coupled provisioning of optical network and IT resources. On the other hand, it is used as a demonstration platform for testing the software prototypes within the project and to demonstrate to the research and business community the project approach and solutions. In this work, we discuss our experience in the deployment of the testbed and share the results and insights learned from our trials in the process. Additionally, the paper highlights the most relevant experiments carried out in the testbed, aimed at the validation of the overall GEYSERS architecture. The current development of cloud computing technologies not only demonstrates an emerging trend towards the integration of cloud based infrastructure services through new inter-cloud and hybrid models, architectures and integration tools. It also supports the provisioning of a common/interoperable environment for migrating existing infrastructures and services to virtualised cloud-based architectures Enhanced Cloud Computing services such as Infrastructure as a Service (IaaS), Platform as a Service (PaaS) or Software as a Service (SaaS) require the dynamic allocation and virtualization of hardware and software elements in order to optimize resource utilization . These requirements stretch the communication network and protocols, especially when cloud facilities (e.g., Data Centres (DCs)) are interconnected through the Internet. Moreover, there is a need to support these services with enhanced network resource provisioning mechanisms in a cost-effective scalable and dynamic way  . Finally, bandwidth requirements  and computing capacity for future applications will stress and exceed current network and IT infrastructure capabilities. Consequently, providers are facing the challenge to adapt their current infrastructure services provisioning models. Figure 9: Resources and roles in Demonstrator 2. As well as Demonstrator 1, an initial version of this demonstrator has been presented during the FuNeMS'12. In this event, we demonstrated the NCP+ procedures for the on-demand setup and tear-down of different types of connectivity services over a single-domain virtual infrastructure. In the unicast service, the traditional GMPLS and Path Calculation Element (PCE)  based functionalities of the NCP+ have been shown. This initial service, where network and IT resources are considered as separate entities and not jointly processed, contrasted with the most efficient scenario of the anycast services also demonstrated. In this case, the selection of the IT end-points is fully delegated to the NCP+ that is able to optimize the overall resource utilization, searching for storages/servers compliant with the description specified in the application requests and, at the same time, reachable through efficient network paths providing the required bandwidth. In this context, we showed the innovative procedures in support of IT advertisement mechanisms to make the NCP+ aware of the capabilities/availabilities in the resources located at the IT end-points, the RSVP-TE signalling extended for the control of the NIPS call and the NIPS path computation at the centralized PCE, with the anycast routing algorithms to jointly select the network and IT resources  .