Post date: Jul 5, 2012 2:46:08 AM
According to [3], with the aim of improving the technology used on space explorations, the main objective of Interplanetary Internet (IPN) is to connect the network on Earth with the network on-board satellites and spacecrafts and the network governing machines deployed in explored planets. Thus, while the internet that we know is a network of 'connected' networks, IPN is a network of 'disconnected' internets.
The initial idea for IPN is to use existing terrestrial internet protocol stack in space. However, it was found out that such technique won't work due to the differenve between the network environments on Earth and in space. Their difference can be captured in three points:
1. The presence of long (speed-of-light) delay in space communication.
2. The presence of intermittent, sometimes unidirectional, connection
3. The presence of disruption in signals
In 2003, Kevin Fall (then affiliated with the Intel Research Lab in Berkeley) introduced (in [2]) the concept of Delay Tolerant Network (DTN), a term used to describe the network environment characterized by the three items above. He pointed out that IPN is only one of the environment capturing the concept of DTN, and such problems are also being experienced in other terrestrial network environments, specifically in mobile ad hoc networks, sensor/actuator networks, etc. The research now is focused on developing a protocol stack that can be used by DTNs. The assumption (based on [1]) for DTN-enabled applications is that they will minimize the roundtrip exchanges and will determine the relative importance of the data being sent, as well as the length of time such data will be useful. Such information will be useful in the scheduling and path selection of the message.
The proposed solution to DTN is an overlay network which will be implemented in any DTN node (regardless of the underlying protocol stack). The major features, based in [1], of such architecture are:
1. Flexible Naming Scheme
- Nodes are identified using a naming syntax (syntax that follows the Uniform Resource Identifier (URI)). The identifier of a DTN node is composed of two parts, the global identifier which is expected to be unique per node and the entity identifier that needs not be unique but must be resolvable in its respective network. It also uses the idea of late binding, thus the name-to-address resolution need not be done in the source node.
2. Postal Service delivery option
The types of services it offer are email-like type. Recall that DTN-enabled applications are not interactive.
3. Variable-length Bundles
Messages (bundles) sent across the network are of variable-length. The reason is that is it expected that this will be helpful in minimizing the number of roundtrip times.
4. The idea of Custody Transfer
There are two ways to assure reliable delivery in the bundle layer. First, an optional end-to-end acknowledgement can be employed. Second, bundle layer may use custody transfer which offers coarse-grained retransmission. Once a node accepts custody responsibility, it promises reliable delivery of the data to the receiver. In this way, the node will not erase the data unless it is assured that another node in the path accepted custody. It also uses the idea of timers, retransmission and acknowledgement. Of same work as that in TCP, but this time, the endpoints are not the source and the destination but two intermediary nodes in the network.
5. The use of Persistent Storage
In case there are no available links in the path, a message must be stored.
6. The use of Administrative Records
Aside from bundles containing message, there are bundles that describe the state of other bundles and bundles that are utilized as signals.
In order to implement the bundle layer, a convergence layer is expected. This layer assures that the underlying network is reliable and facilitates flow and congestion control. This layer may have different implementation depending on the underlying protocol stack.
The research for a suitable architecture remains active and there still are issues that needs to be settled. Some of these, according to [1], involves the idea of reliable multicast, security mechanisms, routing and flow and congestion control at the bundle layer.
In conclusion, most of the techniques done in the classical internet is also adapted for DTN. As can be observed, the overlay network still adapts the store-and-forward mechanism. However, instead of relying on the end-to-end principle, the DTN network is busier because state information is maintained in the core of the network.
References:
[1] Vinton Cerf et al, Delay-Tolerant Networking Architecture, RFC 4838, 2007
[2] Kevin Fall. A Delay-Tolerant Network Architecture for Challenged Internets. In Proceedings of the 2003 conference on Applications, Technologies, Architectures and Protocols for Computer Communications (SIGCOMM '03). ACM, New York, NY, USA, 27-34, 2003
[3]The Interplanetary Internet Project Page: http://www.ipnsig.org/aboutstudy.htm