Delay Tolerant Networking

Abstract

This paper aims to illustrate the issues involved in Delay Tolerant Networking in respect to an Interplanetary Network and uses our local solar system and the missions being conducted within it as references for the issues experienced in a practical working environment. It discusses various techniques in sending data over extreme distances and the limitations of current Internet standards. It explores the bundle layer solution to networking data as well as common coding techniques that create redundancy in an environment that is conducive to packet loss. Additionally it mentions the possible future of an Interplanetary Network.

Problems faced by DTNs

One of the main issues of interplanetary networking is the long delays in transmission. This can be due to a number of factors; Transmitting a signal at the speed of light to Mars takes between 4 minutes to 20 minutes to reach it’s destination. Furthermore, due to orbital mechanics (rotation of planets) there may only be a set time in which transmissions can be sent or received, and no way of communicating outside these time limits.

Protocols such as TCP are perfect for use in areas where there is little or no delay, such as the internet. It is part of the reason why TCP has become almost the de-facto standard for packet transmission in the world wide web and most networks. However, when it comes to areas that do not have a guaranteed connection, or that have a high delay, TCP performs poorly. The Border Gateway Protocol, which is built on TCP, does not perform well when TCP is unable to keep a connection established. Therefore for interplanetary networking, where such delays are not only commonplace, but the rule rather than the exception, TCP is unsuitable as a protocol.

Furthermore, a difficult problem is that of route calculation, especially in networks with scheduled or intermittent connectivity, where network links are created and removed in a predictable way. Normal IP route calculation can be impossible in these situations, as there may not be any way to connect to the destination at all. This is the norm for most interplanetary networks.

A solution to these problems was required, so that communication between planets could be established in some useful way. Delay Tolerant Networking (DTN) was designed to operate in areas (as the name of the protocol suggests) where there may be long delays. Delay tolerant networking has 3 fundamental principles to it:

A postal model of communication
Tiered functionality
Terseness
Postal Model of Communication

The idea of this model is that messages are sent asynchronously. In typical networks that we use every day on this planet, an acknowledgement is generally required for every message that is sent before any other message is sent. In an interplanetary network, there is no guarantee that we will receive an acknowledgement immediately, so new messages are sent without waiting for an acknowledgement for any old ones.

Tiered Functionality

In the same way that TCP functions , with a layer that performs a specific function (data layer, transport layer etc.) a DTN would operate with the same tiered functionality. However, the DTN would have an additional layer called the bundle layer, which performs all the functions necessary for communication in a network similar to that of interplanetary networks. This would function almost like a store-forward layer, but with specific adjustments. These will be discussed in detail later.

Terseness

Generally, DTN are designed to communicate as little as possible, even if that means some extra processing complexity. In interplanetary networks, bandwidth cannot and is not assumed to be cheap, so the less communication required the better.

As an example of a DTN communication, consider this example. If a user were to perform a file upload to a device on Mars from Earth, the DTN would not only send the name of the file to upload, but the username and password needed to initiate the file upload and any other metadata required. It would also do this unprompted. The units of data sent by the DTN are called bundles , and a specific store-forward layer in DTN architecture would need to be created. This would be called the bundle layer.

Store and Forward Routing

Store and Forward message routing is a solution to the problems faced by DTNs such as intermittent connectivity and opportunistic contacts. In this technique messages are sent to a node, which verifies it’s integrity before forwarding it on to another node. If the message can not be sent right away due to lack of connectivity or delays then the message is stored locally until it can be sent. In this way problems of delays, opportunistic contacts or a lack of end to end connectivity are overcome.

Bundle Layer Protocol

DTNs utilise Store and Forward message routing by adding a new protocol layer under the application layer and above the region specific lower layers. This new layer is used across the all regions of the DTN and so applications can communicate over different regions through it. Taking the internet protocol for example we can see how this is done. The new protocol layer called the bundle layer sits between the application layer and the transport layer.

Application Layer
Bundle Layer
Transport Layer
Network Layer
Link Layer
Physical Layer
Bundles

The bundle layer is said to transport bundles (messages). A bundle has following characteristics: the source applications user data, information provided by the source application on how to process the user data and a header given by the bundle layer.

Bundles layers communicate between each other using sessions and don’t require acknowledgements which results in small or no round trip time for messages. How ever this leads to the question of how communication is done reliably.

The lowers layers in part are used for reliability of communication. For example the transport layer will retransmit any segments that are not acknowledged.

Custody Transfers

The bundle layer can do retransmission if instructed to do so by the application layer in the form of a custody transfer. Because there is no end to end connection in a DTN a message can be ensured to get to it’s destination node using custody transfers. In a custody transfer the message is sent to a successive node and requires an acknowledgement within a certain time frame. If the acknowledgement is not received by the bundle layer of the sending node then the message is retransmitted. Once the message has been acknowledged then the custody of that message has been given to that node which becomes the custodian for that message. The new custodian in turn tries to move the message to a successive node which is closer to the destination. A return receipt can be requested by the original sender which is a confirmation from the destination that the message has arrived and this ensures the communication. Custody transfers and reliable transport layer protocols advance the message closer to it’s destination. Retransmission can only occur between successive nodes and so the advance of retransmission points reduces the retransmission hops and consequently the network load caused by them.

Security

Nodes in a DTN check the integrity of the message that go through them. Part of this check includes authenticating the sender information. When a node receives a message it authenticates the sender’s signature and then replaces it with it’s own signature before sending it on to the next node.

Data Coding and Compression

One of the main issues of networking in extreme conditions is data integrity. It is to be expected that faint transmissions from spacecraft many millions of miles away will be subject to cosmic radiation (like communication satellites in LEO experience during solar flare activity), interplanetary media and occultation’s by planets, sun or asteroid. To counter these issues a mission data stream is augmented by carefully designed convolution codes to enhance redundancy. The convolution codes were used first for the Galileo and Voyager programs and were combined with a Reed-Soloman coding. The complexity of these codes allows for a certain amount of data to be lost during transmission but still be able to recover the data through the decoding process. On the flip side data compression, which is the removal of intrinsic redundancy from a data stream, has been used for compressing images which in conjunction with coding allowed for a power of two increase in data return in real world tests.

Mobile Internet Protocol

NASA experimented with the mobile form of IP for the first time onboard the ill-fated Columbia mission of January 2003. The main advantage that they were exploring was the ability of the successive handing off of communications between the shuttle and NASA’s satellites and ground stations. The shuttle was able to transfer multiple files via udp/ip over a mobile IP link which changed base stations several times whilst traveling at approximately 17000 mph.

UDP was chosen so that the regular TCP “chatter” of ACK could be avoided. Instead a NAK command was sent after a specific delay period asking for particular packets to be resent.

Future of the Deep Space Network

Currently the main issue with the Deep Space Network (DSN) is that the southern hemisphere has only one station available (in Australia) to receive signals from interplanetary craft. This leads to the store and forward approach currently in operation by NASA on the majority of its craft. The alternative is to create a solar system wide relaying network using the existing spacecraft in use by NASA. For example, it is hoped that the various space agencies send a series of micro-satellites to Mars for relaying information to a main communications satellite (similar to the currently cancelled Mars Telecommunications Orbiter (MTO)). Onboard the satellite will be the laser communication platform that will be used as a router to facilitate a the high-speed data link to Earth. Currently all communications are sent by radio waves back to Earth, however these signals are aimed at a broad area of space and wastes a majority of the original signal strength. By using a focused laser it will allow for the signal strength received on Earth to be far stronger and therefore allow for less requirements for coding and higher amounts of compression. It is estimated that data transmission from Mars will increase by a factor of ten whilst also allowing for greater bandwidth availability for future missions such as the Mars Sample Return Mission, several other rovers that are in development and the planned human landing on Mars for around 2030.

References and Definitions

http://www.dtnrg.org/wiki
http://ipinspace.gsfc.nasa.gov – Mobile IP In Space Experiment
http://www.spaceref.com/images/ipn/ipn_dns_ani.gif – Example of DSN use
http://deepspace.jpl.nasa.gov/technology/95_20/return.htm – Information about convolution coding and compression.
http://ipnsig.org – Interplanetary Internet Project
http://www.dtnrg.org/papers/ieee-comsoc-article.pdf – Delay-Tolerant Networking – An Approach to InterPlanetary Internet
http://www.ipnsig.org/reports/IAF-Oct-2002.pdf – The Interplanetary Internet: A Communications Infrastructure for Mars Exploration
http://www.ipnsig.org/reports/SpaceOps-Oct-2002.pdf – Towards an Interplanetary Internet: A Proposed Strategy for Standardization
http://www.ipnsig.org/reports/DTN_Tutorial11.pdf – DTN Tutorial
http://www.ipnsig.org/reports/tutorial/index.htm – The Interplanetary Internet: Architecture and Key Technical Concepts
http://www.ipnsig.org/reports/Lesh-IPN-Technologies.pdf – Technologies for the Interplanetary Network
LEO = Low Earth Orbit. Between 200 – 2000 km (124 – 1240 miles) above the Earth’s Surface.