Report from the Workshop

General Information

The 2011 ACM SIGCOMM ICN workshop do place on Friday, August 19th in Toronto. Over 80 people registered to the workshop.

The workshop received 41 submissions from Asia, Europe, and North America (China, Finland, France, Germany, Greece, Ireland, Israel, Italy, Netherlands, Norway, Sweden, South-Korea, Switzerland, UK, USA), out of which 12 papers have been accepted for the final program -- the acceptance rate was 29%. The distribution of accepted papers' first authors' countries was as follows:

Best Papers

• Ali Ghodsi, Teemu Koponen, Jarno Rajahalme, Pasi Sarolahti and Scott Shenker; Naming in Content-Oriented Architectures
• Diego Perino and Matteo Varvello; A Reality Check for Content Centric Networking

Panel Discussion

• Bengt Ahlgren, SICS, Sweden
• Van Jacobson, PARC, USA
• Teemu Koponen, ICSI, USA
• Ramesh Sitaraman, University of Massachusetts, Amherst & Akamai, USA
• Dirk Trossen, University of Cambridge, UK

The panelists gave short presentations with position statements that led into an interesting discussion. The summary of the statements (in the order they were given at the workshop and as noted by the organizers):

Van Jacobson, PARC, USA

• Want to evolve IP;
• overcoming infrastructure dependencies;
• don't need a topology as a requirement to operate;
• do everything that IP does; and
• possibly run DONA/PSIRP/NetInf over CCN.

Teemu Koponen, ICSI, USA

• Different types of research topics for ICN;
• some are less well understood, but highly critical;
• need less attention for: HTTP -- works OK, just need to decide whether to use PKI or self-certification;
• more attention for: privacy, inter-domain routing;
• more attention for: many objects -- FIBs can't be huge and fast -- aggregation implies hierarchical topological names;
• with larger objects, slow huge FIB, but need faster packet delivery -- requiring location addresses;
• immediate attention for: ubiquitous caching -- which did never really work (cf. history of web caching);
• example Facebook: CDN image caching considered inefficient;
• unclear whether we should really replace the whole infrastructure for logarithmic cache hit rate.

Dirk Trossen, University of Cambridge, UK

• ICN hour class waist -- what could it look like?
• proposing marriage between information and computation;
• working on the hour glass abstraction is key piece;
• need to embed larger community into ICN research (e.g., database community)

Bengt Ahlgren, SICS, Sweden

• Copyright issues are important to consider in ICN;
• there is less control for content owners;
• we may need technologies and laws to satisfy content owners;
• access/eyeball counting required;
• cache management -- collaborative caching will be useful;
• .. with combined cache and traffic management; and
• considering congestion control, and supporting DTN scenarios.

Ramesh Sitaraman, University of Massachusetts, Amherst & Akamai, USA

• Adopting clean-slate ICN architecture to be adopted in real world?
• 5 business drivers for CDN today: availability, performance, scalability, cost and security;
• scalability: "middle mile bottleneck";
• cost: exponential decrease of bandwidth cost;
• energy cost soon will be dominant;
• if ICN should fly, it has to meet current CDN availability requirements!