Architectures and resource management for WAN cloud networking

点击翻译按钮获取中文摘要

Cloud computing services are expected to grow annually 25% in the period 2012-2016 (IDC forecast). In fact this growth is five times bigger than the overall growth of IT industry, which indicates a major technological shift. Also, public cloud services constitute a significant part of enterprise spending and it is expected that this spending will reach $207 billion by 2016 (Gartners’s IT spending report). Public cloud computing is increasingly relying on WAN networking since the cloud resources can be located quite far from the users. Traditional internet has many limitations and in order to support cloud computing, a cloud support network should be significantly more agile and cost effective. Network as a Service (NaaS) is an important new networking paradigm that can help in achieving these cloud networking objectives. While NaaS has many definitions, in this project we consider NaaS as a service that is offered to users (mostly enterprise users) by a virtual network operator (located in a cloud) in order to manage and optimize WAN performance perceived by the user’s applications. In particular the focus is on latency, reliability and security. In general NaaS can be offered in combination with other cloud services or as a separate service. The importance of NaaS concept is validated by the fact that Cisco recently spent $1.2 billion to acquire Meraki - a leader in NaaS for small and medium companies. Note that while NaaS provides a framework that can facilitate WAN cloud networking it does not solve by itself the limitations of traditional Internet but rather allows creating better architectures and mechanisms build on top of the Internet that can address several issues. Therefore, apart optimization of NaaS for current Internet, it is also important to study which future Internet architectures are fitted best for cloud networking.In this project we propose to develop resource management mechanisms for WAN cloud networking that are based on different NaaS architectures including architectures build on top of future Internet and involving the SDN (Software Defined Networking) concept. These algorithms will be build based on an economic framework that can integrate many objectives of cloud networking. Apart giving solutions for particular NaaS architectures, the objective is to compare the performance to see which NaaS architectures and which future Internet options are most promising for cloud networking. We believe that the project results will be useful in two dimensions. First, the derived models and algorithms for NaaS can help virtual network operators to improve performance of cloud networking. Second, the study of the future Internet options in the context of cloud networking can be useful to decision makers that have a say in the future Internet developments.