NETCOM Research Group

Cisco estimates that by 2019 traffic will be 64x than in 2005. Network caching can help cope with this traffic deluge as, by replicating content closer to the users, it reduces bandwidth consumption. For this reason, its role in 5G is considered crucial by many authors. Nonetheless, current caching techniques provide limited benefits in terms of (i) operational cost reduction for Internet Service Providers (ISPs), (ii) user experience improvement in video delivery and (iii) encrypted content transmission. These limitations are particularly relevant given that (i) ISP revenue has decreased in recent years (source: ITU), (ii) video traffic will represent 80% of all traffic (source: Cisco) and (iii) most of the traffic is now encrypted. However, most of the literature on caching is mainly focused on improving the hit ratio, which is interpreted as the most important indicator of cache performance. As a consequence, the aforementioned limitations have not been sufficiently addressed. On the contrary, Andrea Araldo will show that optimizing the metrics that really matter for ISPs (cost) and users (video quality) often conflicts with hit ratio maximization and proposes caching policies that successfully optimize the relevant objectives. Moreover, regarding (2), while classic caching assume that an object is represented by only one file, the same video can be represented at different bitrates and resolutions, which require different storage space and bandwidth, but also result in a different quality perceived by the users. Novel caching techniques that explicitly choose the representation to replicate will be presented. As for (iii), we observe that classic caching techniques assume that the cache owner, e.g., the ISP, has a complete visibility of the content being served. This does not hold anymore, since Content Providers (CPs) want to have exclusive control over their content and they obfuscate it, by means of HTTPs. Therefore, ISPs have often to cope with huge flows of data that they cannot reduce, since caching is not applicable, whence frequent and sometimes famous disputes between ISPs and CPs. Content Delivery Networks (CDNs) have not really solved the problem, as they require CPs to lose again the exclusive control over their content, since it must be transferred to the CDNs. A. Araldo will present Content Oblivious Caching, an architectural and algorithmic solution that consents to ISPs to cache even if they cannot see the content being served. Content Oblivious Caching is an example of how intriguingly inter-dependent are technical and economic aspects, when designing a solution in the content distribution context.

The methods used to analyze and support the new solutions are optimization models, stochastic control and simulation.

Andrea Araldo will also briefly present his current research activity on Intelligent Transportation Systems, concerning "Sustainable Travel Incentives with Prediction, Optimization and Personalization".

The content of this talk is covered mainly in these three publications:

• A. Araldo, D. Rossi, F. Martignon, “Cost-aware caching: Caching more (costly items) for less (ISPs operational expenditures)”, IEEE Transactions on Parallel and Distributed Systems (TPDS), 27(5), May 2016
• A. Araldo, F. Martignon, D. Rossi, “Representation Selection Problem: Optimizing Video Delivery through Caching”, in IFIP Networking, 2016
• A.Araldo, G. Dan, D. Rossi, “Stochastic Dynamic Cache Partitioning for Encrypted Content Delivery”, in ITC, 2016

About Andrea Araldo

Andrea Araldo is currently a postdoctoral researcher at the Massachusetts Institute of Technology (MIT), where he works at the Intelligent Transportation Systems Lab. His current research interests are Network Optimization, Intelligent Transportation Systems, Content Distribution in the Internet, SDN, ICN, Network Measurements. He received his BSc and MSc in Computer Engineering from Universita' di Catania (Italy) and his PhD in Computer Networks in 2016 from Universite' ParisSud and Telecom ParisTech (Paris), which he also joined as postdoc. He worked for the Italian academic research institute of networks (CNIT) and in the FP7 EU Projects Ofelia and mPlane. He was visiting researcher at KTH Royal Institute of Technology (Sweden).

This event will be conducted in English