An Architecture for Cache Consistency Support in Information Centric Networking Zhen Feng, Mingwei Xu, Yu Wang and Qing Li Dept. of Computer Science and Technology Tsinghua University, Beijing, China feezhen@sina.com, xmw@cernet.edu.cn, wangyulj@sina.com, andyliqing@gmail.com Abstract—Information centric networking is proposed to deal with the inefficiency of content distributions in traditional net- works. It replaces the current host-centric communication para- digm with the content-centric one, so all the network devices can identify and cache the contents passed by. Traditional cache consistency approaches which rely on the origin server to validate cached contents will not be practical in ICN. The origin server might be overwhelmed by the large vo- lume of validation requests from ubiquitous caches. In this paper, we propose a cache consistency architecture named Content Update Validation System (CUVS). It makes use of the servers located in each domain to provide the content vali- dation service. It relieves the origin server and the routing proto- cols from caring about the cache consistency issue. The system is designed as a hierarchical overlay network in ICN, so it is easy to deploy and scales well. Keywords—information-centric networking; named data networking; cache consistency I. INTRODUCTION The usage of the Internet has changed dramatically since it was invented in the late 1960s. The initial role of the Internet was designed to share the then expensive hardware and com- puting resources. However, following the advent of the World Wide Web, the Internet has gradually evolved into an informa- tion repository. Despite the variations of its functionality, the architecture of the Internet remains unchanged. Information Centric Networking (ICN) is an architectural redesign proposed to deal with the most prominent require- ments nowadays, e.g. efficient content distribution, mobility, security, etc. Many schemes have been proposed, including NDN [1, 2], DONA [3], NetInf [4], and PSIRP [5], etc [6]. Although these schemes differ in technical details, the feature of ubiquitous caching is widely adopted. When caching is employed, there will be many copies of the same content in the network. As a result, when the origin copy of content changes sometime, the cache consistency among all the copies needs to be taken into consideration. For contents which will never be modified after publishing (e.g. most mov- ies), the cache consistency is not an issue. For contents which will be updated in real time (e.g. the stock index), the cache consistency is not needed either since the contents will not be cached at all. But, for those contents which may change occa- sionally (e.g. software updates and blog updates), the best strategy is caching them while incorporating the cache consis- tency mechanism at the same time. Caching consistency is not a recently emerged issue. Many cache consistency schemes have been developed in the follow- ing fields: 1) file system and distributed database system; 2) DNS; and 3) web caching. However, the scenario in ICN is quite different from those scenarios above. First, strong consistency is normally used in the file system or distributed database. It will be disastrous and cannot be tolerated if stale data is returned from a financial database no matter how good the performance the cache can provide. However, most of us are willing to download a software patch from a cache even if it may be stale in the last 5 minutes. The rational includes: 1) the probability of staleness is relatively little; 2) no serious consequences will be incurred in case of staleness; and 3) the cache may provide far better performance compared with the origin server. Second, weak consistency is widely used in DNS since the change of the IP address of a domain is infrequent. The TTL of a DNS record is normally assigned to hours or days. But the change frequency of contents in ICN is different from the DNS records. They are more heterogenous and largely depend on the content types in ICN. Moreover, the cost of getting a new copy of content is much higher than a new DNS record. Third, web caching behaves like a content-oriented system and shares many similarities with ICN. The main differences between web caching and ICN stems from their scale. There are usually a small number of web caches in an ISP domain, so they can be manually configured into certain topology to maximize the performance. Meanwhile, delicate protocols can be implemented to facilitate their collaboration. These advan- tages make the cache consistency issue of web caching much easier. In the ICN, however, all the network equipments may cache contents. Their topology reflects the status of network- ing rather than the need for collaboration. The routers in ICN will run routing protocols as well, but we think the main goal of routing is to find the requested version of content rather than checking whether the content is stale or not. Also, it is inefficient to implement the cache consistency mechanism in the routing protocols. As far as we know, no academic efforts have been put on the cache consistency issue in the ICN field. In our opinion, the wide spread caching in ICN makes the issue more intracta- ble, and the traditional approaches which mainly aim to oper- ate in a small size or relatively self-managed environment (e.g. web caching and CDN) cannot be efficiently applied in ICN. Globecom 2013 - Next Generation Networking Symposium 978-1-4799-1353-4/13/$31.00 ©2013 IEEE 2126