MASTER: sMart mAnagement for futuriStic daTa cEnteRs

近年来，云计算已被证明是托管大规模互联网应用程序的理想平台。因此，迫切需要进一步开发云基础设施和服务，并提高其敏捷性、成本效益、性能和可用性。网络功能虚拟化（NFV）和软件定义网络（SDN）等新兴技术很有希望满足这一需求，因为它们允许动态构建服务作为虚拟网络功能（VNF）链，这些虚拟网络功能可以使用SDN功能轻松互连。为未来基础设施提供高效、动态的云管理框架，能够满足大型在性能和可用性方面扩展Internet应用程序，同时最大限度地降低运营成本。这需要设计新颖的资源分配方案来管理基于NFV的服务，并设计能够利用SDN灵活性而不产生其潜在限制的云架构。具体来说，我们打算研究以下关键挑战：（1）基于VNF的云服务的设计和管理：随着虚拟化技术的成功，最近的提议提倡虚拟化网络功能（例如，防火墙、路由），并将它们作为VNF提供以构建互联网服务。在这种情况下，我们打算解决与基于NFV的服务的动态配置相关的挑战，以确保其性能并降低运营成本。(2)云中的可感知性管理：当今的云计算用户期望严格保证其服务的可用性。然而，目前还没有准确和有效的技术来评估在云上运行的服务的可用性。因此，我们计划开发能够评估服务可用性的模型，并利用它们来设计新一代的可用性感知资源分配方案，满足所需的可用性，同时最大限度地减少所产生的成本。(3)基于SDN的云计算架构的设计和管理：SDN技术最近已经成为一种新的范例，提供了动态配置云网络所需的可编程性。尽管其优点，传统的SDN架构遭受可扩展性的限制。因此，我们打算研究与设计一个新的基于SDN的架构，克服这些限制，同时保持SDN的好处相关联的挑战。该项目的成果预计将在开发新的架构和管理框架的下一代云基础设施和服务作出重大贡献。该项目也是研究生获得NFV和SDN等技术专业知识的绝佳机会，这些技术在工业和学术界获得了巨大的发展势头。

In the recent years, cloud computing has proved to be the ideal platform for hosting large-scale Internet applications. As a result, there is a pressing need to further develop cloud infrastructures and services and enhance their agility, cost-effectiveness, performance and availability. Emergent technologies like Network Function Virtualization (NFV) and Software Defined Networking (SDN) hold great promise to fulfill this need as they allow to dynamically build services as chains of Virtual Network Functions (VNFs) that are easily inter-connected using SDN capabilities.This project aims at designing scalable, efficient and dynamic cloud management frameworks for futuristic infrastructures able to satisfy requirements of large-scale Internet applications in terms of performance and availability while minimizing operational costs. This requires devising novel resource allocation schemes to manage NFV-based services and designing cloud architectures able to leverage the flexibility of SDN without incurring its potential limitations. Specifically, we intend to investigate the following key challenges:(1) Design and Management of VNF-based Cloud Services: with the success of virtualization technology, recent proposals have advocated to virtualize network functions (e.g., firewalls, routing) and offer them as VNFs to build Internet services. In this context, we intend to address challenges related to the dynamic provisioning of NFV-based services with the goal of ensuring their performance and reducing operational costs.(2) Availability-aware management in the Cloud: today's cloud users are expecting stringent guarantees on the availability of their services. However, there are currently no accurate and efficient techniques to evaluate the availability of a service running on the cloud. We therefore plan to develop models able to evaluate service availability and leverage them to design a new generation of availability-aware resource allocation schemes that satisfy the desired availability while minimizing the incurred costs.(3) Design and Management of SDN-based Cloud Infrastructures: SDN technology has recently emerged as a new paradigm offering the programmability required to dynamically configure cloud networks. Despite its benefits, the traditional SDN architecture suffers from scalability limitations. We therefore intend to investigate the challenges associated with the design of a novel SDN-based architecture that overcomes these limitations while keeping SDN benefits.The outcome of this project is expected to make significant contributions in developing novel architectures and management frameworks for next-generation cloud infrastructures and services. This project is also an excellent opportunity for graduate students to acquire expertise in technologies like NFV and SDN which are gaining significant momentum in industry and academia.