Survivable, Latency-Aware, Delay-Tolerant, and Disruption-Tolerant Mobile Wireless Networking

Survivable, delay-tolerant, and disruption-tolerant communication benefits from new thinking at multiple levels, from new routing and forwarding paradigms at the network layer, end-to-end mechanisms supporting lossy channels at the transport layer, to applications that are latency-aware, delay- and disruption-tolerant.

Three research projects are described on this page that begin to address these issues: SUMOWIN at the network layer, ETEN at the transport layer, and WVM at the application layer.

Presentations

James P.G. Sterbenz and Rajesh Krishnan, “Survivable and Disruption Tolerant Networking: Issues, Challenges, and Research Directions”,
[ full size and color for display | 2-up limited color for printing ]

This is a longer, combined, tutorial presentaiton that includes most of the three presentations below.


Survivable Mobile Wireless Networking (SUMOWIN)

The SUMOWIN project has roots in assisting in Rob Ruth's effort to put together a Mobile Wireless Information Assurance DARPA program in 1999 as a follow-on to the GloMo program, for which he was program manager. Doug Maughan carried the torch forward and, and formed the DARPA SUMOWIN (Survivable Mobile Wireless Networking) seedling program at DARPA in 2000. This work has also served as the basis for input to the NSF Wireless Network Workshop Final Report in 2003.

The primary goal of the SUMOWIN program is to ensure that networks are resilient in the face of:

It is important to note that fault tolerance assumes random failures, whereas network survivability must be resistant to coordinated attacks by an intelligent adversary.

We performed a set of tasks to help define the program, consisting of:

  1. Survivable topological connectivity and low probability of detection in ad hoc wireless networks
  2. Automated selection of network participants using topological, geographical, and spatial techniques
  3. Survivable key management for wireless networks
  4. Public key infrastructure capabilities and wireless network security
  5. Wireless network recovery and reconstitution techniques
  6. The role of satellites in future survivable, mobile wireless networks
  7. Communications in the military battlefield

One of the key ideas to come from this work is the need for new routing and forwarding paradigms that do not assume that a stable (bi-directional) end-to-end path can ever exist. This is needed for conventional routing algorithms to converge, and is called eventual stability by the distributed computing community.

Network and communication mechanisms should expect weak- intermittent- and episodic- connectivity as part of their normal operation. Furthermore, these mechanisms should expect and exploit mobility to enhance survivability.

We observe that challenging communication environments can benefit by communication using the eventual connectivity paradigm, in which information is transferred as far as possible, whenever possible, in the direction most promising to reach its destination. This means that when a stable end-to-end path is not available, nodes store-and-forward information, switching bursts of data as individual links become usable (wireless burst switching). Furthermore, information can be physically carried by a mobile node (store-and-haul) through a region where communication is not possible, due to interference or the fear of eavesdropping.

We also observe that communication protocols should not rely on strongly-connected symmetric paths for operation, but should use open-loop control (rate control and FEC) when needed to allow communication even when a reliable back-channel is not available. This is commonly the case when mobile nodes do not have the same transmission power, or when a node wishes to receive information but remain radio-silent.

Publications and Presentations

The research issues for network survivability were published as:

James P.G. Sterbenz, Rajesh Krishnan, Regina Rosales Hain, Alden W. Jackson, David Levin, Ram Ramanathan, and John Zao,
“Survivable Mobile Wireless Networks: Issues, Challenges, and Research Directions”,
International Conference on Mobile Computing and Networking (MobiCom),
Proceedings of the ACM Workshop on Wireless Security (WiSe 2002),
Atlanta, GA, USA, 28 September 2002, pp. 31–40.
[ abstract | PDF ]

An extended version of the WiSe 2002 presentation is available [ PDF ]

Research Staff

James P.G. Sterbenz (Principal Investigator),
Jerry Burchfiel, Regina Rosales Hain, Alden W. Jackson, Stephen Kent, Rajesh Krishnan, David Levin, John Lowry, David Mankins, Ram Ramanathan, Gregory D. Troxel, Ron Watro, William Watson, and John Zao.
[ BBN SUMOWIN project page ]


ETEN: Explicit Transport Error Notification

It has long been known that there are benefits in separating flow, congestion, and error control mechanisms to enhance performance. More recently, it has become clear that it is beneficial in wireless networks to convey some information about the communication channel and the reason for losses to higher layer protocols.

Protocols such as TCP that combine flow/congestion/error control mechanisms can perform very poorly in the wireless environment when they react incorrectly to a loss event. TCP assumes that the lack of an ACK is due to congestion (or congestion avoidance) and throttles the sender. If the loss is due to a channel error under light load, this is the wrong behaviour; the source should retransmit as soon as possible.

This project explores the benefits to distinguishing channel errors from congestion to end-to-end protocols, with a focus on TCP mechanisms. ETEN (explicit transport error notification) is a type of explicit loss notification that we explored, and is the is the error-based analog to ECN (explicit congestion control). It is important to note that ETEN cannot be derived from ECN or vice-vera; both must be explicitly conveyed. The reason for this is that a particular packet may first contribute to congestion, and then be lost, and thus there is no general analytical relationship between P[channel error] and P[congestion].

Publications and Presentations

Rajesh Krishnan, Mark Allman, Craig Partridge, and James P.G. Sterbenz,
Explicit Transport Error Notification for Error-Prone Wireless and Satellite Networks,
BBN Technical Report 8333, 7 February 2002 (revised 22 March 2002).
[ abstract | PDF ]

Rajesh Krishnan, Mark Allman, Craig Partridge, James P.G. Sterbenz, and William Ivancic,
“Explicit Transport Error Notification (ETEN) for Error-Prone Wireless and Satellite Networks – Summary”,
Earth Science Technology Conference 2002,
Pasadena, CA, USA, June 11-13, 2002.
[ PDF ]

Research Staff

James P.G. Sterbenz (Principal Investigator), Rajesh Krishnan (Co-PI),
Craig Partridge, Mark Allman.
[ BBN ETEN project page ] – includes ns-2 simulation models and software


Latency-Aware (Delay-Tolerant) Information Access

We have also explored the benefits of applications being aware of the underlying network, conveying this information to the user, and allowing the user to alter the behaviour of the application to better serve her needs.

The goal of the Web Vade Mecum (Web constant companion, from Latin) project is to make the Web substantially more usable in environments with unpredictably long latencies to the user, and under episodic connectivity. This work is specifically targetted toward mobile wireless clients, but is also useful in the context of congested networks and servers, as well as with the extremely long latencies of the Interplanetary Internet.

We have prototyped a latency-aware web browser, which provides response time estimates to the user, and allows the user to determine the appropriate fetch behaviour based on these estimates. Additionally, during periods of high connectivity, the browser hoards content based on a user profile graph, so that the probability of local cache hits is maximised during periods when the client is not connected to the Internet.

WVM screenshot

This screenshot shows some of the main features of the WVM prototype. URLs are tagged by color (or icon) to give the user an idea of the estimated response time and freshness. Green links are fast (either recently cached or will have low response time to retrieve), yellow links are cached but old, and red links are not cached and will have long response times. Detailed information on the strength of the network connection and estimated response time per URL are shown in the bottom status bar on mouse-over for individual links. The right-click menus have been modified to allow the user to determine how pages are loaded, given the response time and freshness information: open cached in this window, fetch and open definitive copy in this window, open definitive in a new window (which is non-blocking and allows the user to continue browsing in the current window), and open cached in this window and the definitive in a new window when it arrives.

Publications and Presentations

James P.G. Sterbenz, Tushar Saxena, and Rajesh Krishnan,
Latency-Aware Information Access with User-Directed Fetch Behaviour for Weakly-Connected Mobile Wireless Clients,
BBN Technical Report 8340, 9 May 2002.
[ abstract | PDF ]

An presentaion of the WVM project is available [ PDF ]

Prototype Software at BBN

WVM software block diagram

Research Staff

James P.G. Sterbenz (Principal Investigator), Tushar Saxena (Co-PI), Rajesh Krishnan.
[ BBN WVM project page ] – includes prototype source and installation instructions


Links to Related SUMOWIN and DTN projects


Last updated 9 February 2005 – Valid XHTML 1.1Lynx inspectedW3C A Conformance
©2003–2005 James P.G. Sterbenz <jpgs@sterbenz.org>