FIA Principal Investigators' Meeting

November 14-15, 2013
UCSD, San Diego, Ca





Plan of meeting:
In contrast to previous meetings that have had one or two themes, the planning group decided to organize this meeting around four topics, each of which will occupy approximately a half day. The topics are as follows.

1) Management of faults, which includes detection, localization, isolation/avoidance/working around them, and repair/recovery. This is an aspect of security and also management, and was selected because it seems focused enough for discussion. Many of the FIA proposals have something to say about fault management, but (as with routing) the points of view may be very different. Schemes that include some sort of explicit user choice over routes must be able to feed back failure information (e.g. location of the fault) to the element making the choices.
Schemes that do implicit path selection (like NDN) must be able to detect and deal with problems internally. (In ICNs like NDN, flawed data is presumably a relevant class of fault.)

2) Routing. The question of interest here is not the complete details of how the schemes are currently doing routing, but what the different approaches provide as foundations for routing, and what they require from a routing protocol. Issues of mobility, ICNs, storage in the net, etc. have important implications for what routing must do. The contrast between proposals with a "source route" flavor and those with a more traditional IGP/EGP approach may be revealing, in terms of the required architecture support. Support for mechanisms of the "anycast" nature could be discussed

3) The Internet of Things. Several of the projects have a point of view about sensors, M2M, etc. This topic might include M2C (machine to cloud) issues, economics of supporting small things, and so on.

4) Performance. Most of the FIA proposals do not distinguish themselves from the current Internet on the basis of simple performance improvements (e.g., efficient use of links). However, the different schemes raise different and interesting questions about aspects of performance. Metrics and measures related to mobility might be interesting. The degree to which the performance of the system degrades (or not) under attack (e.g., DDoS attacks) might be interesting. A question is the different balance between data plane and control plane in the different designs, and performance may be a good case study for that question. As with the issue of dealing with faults, designs that give the user some choice must have means to allow the user to evaluate and pick options.

Day 1 - Thursday, November 14

Session 1: Fault management

Jonathan Smith, Nebula Slides
John Byers, XIA Slides
Ray Raychaudhuri, MF Slides
Beichuan Zhang, NDN Slides

Session 2: Routing

Arun Venkataramani, MF
Peter Steenkiste, XIA
Lan Wang, NDN Slides
Jonathan Smith, Nebula Slides

Day 2 -Friday, November 15

Session 3: Internet of Things

Jeff Burke, NDN Slides
Jonathan Smith, Nebula Slides
Richard Martin, MF Slides

Session 4: Performance
Srini Seshan, XIA Slides
Jim Kurose, MF Slides
Rudra Dutta, ChoiceNet Slides
Hakim Weatherspoon, Nebula
William Lehr Nebula
Patrick Crowley, NDN


Hosted by the Computer Science and Artificial Intelligence Laboratory at MIT under agreement with NSF.
Contact David Clark <ddc@csail.mit.edu> with questions or comments.
Content of this page extracted from NSF FIA solicitation and press release. See links above for full text.