Using SDN and NFV to enhance request rerouting in ISP-CDN collaborations

详细信息    查看全文

• 作者：Jeffrey Lai^a ; ^{jeff@jeffl.ai}Author Vitae ; Qiang Fu^a ; ^{qiang.fu@ecs.vuw.ac.nz}Author Vitae ; Tim Moors ; ^b ; ^{t.moors@unsw.edu.au}Author Vitae
• 关键词：CDN ; SDN ; NFV ; QOS ; DNS ; Traffic engineering
• 刊名：Computer Networks
• 出版年：2017
• 出版时间：11 February 2017
• 年：2017
• 卷：113
• 期：Complete
• 页码：176-187
• 全文大小：1116 K
• 卷排序：113

文摘

With the rapid increase of data consumption across residential networks, Content Delivery Networks (CDN) have correspondingly increased in number and size. Similarly, residential Internet Service Providers (ISP) have begun collaborating with CDNs in order to provide better Quality of Service (QoS) to their customers. With this collaboration there is an opportunity for a better request redirection mechanism that connects end users to CDN surrogates. While current request redirection mechanisms that utilize the Domain Name System (DNS) are capable of providing adequate information to users, the granularity at which this information can be updated is limited by the TTL in their DNS responses. In addition to this, the presence of DNS caches located between the end user and their recursive DNS server (Such as in their browser, operating system, or home router) can further increase the time it takes for a DNS based redirection mechanism to provide an end user with optimal CDN surrogate server addresses. While DNS caches generally improve user QoS by decreasing the number of redundant DNS requests that an end user may make, their presence may actually degrade user QoS by not connecting them to a more optimal CDN surrogate when one is available, possibly leaving them connected to a CDN surrogate with low QoS. To work around this limitation imposed by client side DNS caching, we propose a novel approach where an ISP in an ISP-CDN collaboration actively redirects end users to selected CDN surrogate servers via a combination of Software Defined Networking (SDN) and Network Functions Virtualization (NFV) techniques. We present two different approaches that can be used to realize our proposed architecture, both of which have been shown to vastly improve perceived user QoS by several orders of magnitude.