For the PDF version of this feature, click here.

Network-centric warfare today relies upon a mix of wired and wireless networks to deliver mission-critical information. Commanders and tactical warfighters depend on timely information to make informed strategic and tactical decisions on the battlefield.

Network conditions within the tactical infosphere, however, can suffer from extreme conditions, including limited bandwidth, high latency, fading, interference or jamming. Often, the result will be lost data packets and poor communications performance that can impair the ability to share mission-critical information among tactical operations centers and individual warfighters.

DF Raptor, Digital Fountain's patented FEC technology, addresses the problems posed by such severe communications conditions. Adopted as a global standard for the wireless distribution of multimedia content over 3G cellular networks, DF Raptor provides the means to improve the reliability of tactical communications links, guarantee faster delivery of critical data over today's networks, and allow efficient and deterministic control of network utilization.

Using DF Raptor, systems integrators can develop applications that provide fully reliable and interoperable communications to protect critical mission data over any type of channel under extremely adverse network conditions. Such applications can help individual tactical units share real-time information and improve battlefield situational awareness and tactical performance.

In particular, DF Raptor makes possible the efficient and reliable delivery of data files and streams over unicast or multicast packet networks. Military communications architectures that can especially benefit from DF Raptor include:

• Point-to-point data transfer over unicast networks to support bulk data transfers;

• 1-to-N data transfer over multicast networks to support data distribution and dissemination; and

• N-to-1 data transfer, an innovative potential use of DF Raptor to optimize the delivery of data by using more than one transmitter or channel to transfer data to each receiver.

Unlike FEC technologies that correct bit errors, DF Raptor recovers data packets that have been “lost” in transmission. DF Raptor has been designed and optimized for powerful packet loss protection, flexibility and low complexity.

DF Raptor is a packet-level erasure code, protecting against the loss of packets that may occur in transmitting data across a packet-switched network. Independent of whether the source data is encrypted, DF Raptor works by encoding and transmitting supplemental data in addition to the original file or data stream. At the receiver, packet loss may prevent the reception of all the transmitted data. However, if enough packets are successfully received, then the DF Raptor decoder can recover the original data, and the number of packets needed for the recovery is only slightly greater than the amount of original data. Furthermore, the original data can be recovered without regard to specific packets or the order in which they are received.

DF Raptor is a fountain code, capable of generating a potentially limitless amount of encoded data from any original set of source data. Such codes are “rateless,” allowing the actual amount of encoded data and thus the code rate to be determined as needed to combat network packet losses. DF Raptor thus makes possible system designs that adaptively adjust to protect against the actual amount of experienced packet loss.

DF Raptor provides exceptional flexibility, permitting system integrators to directly address the needs of specific applications. For example, DF Raptor's fountain characteristic allows DF Raptor-encoded symbols to be continually generated and transmitted until enough have been received to fully recover the original data, at which time the receiver can transmit an acknowledgment to the sender requesting that transmission cease or that the next source block be transmitted. With this “squelch” approach, the overhead associated with FEC coding can be held to a minimum yet provide complete protection against any level of packet loss. Alternatively, if bandwidth is limited or if transmission time must be restricted, then DF Raptor can be configured to generate a finite number of encoded symbols, thereby providing protection against a specific maximum level of packet loss.

Finally, DF Raptor employs exceptionally fast encoding and decoding algorithms that approach the performance limit established by information theory. DF Raptor's encoding and decoding processes operate at nearly symmetric speeds that grow only linearly with the amount of source data to be processed. The processing requirements per unit data are thus constant and, moreover, independent of the amount of packet loss. Unlike bit-level FEC algorithms, hardware support is not required, and DF Raptor can readily be implemented as a scalable, lightweight software-based solution operating on a general-purpose processor, as shown in Figure 1.

DF Raptor provides the means to deliver perishable data efficiently and reliably, either as a complete data file or as a real-time data stream.

Raw and analyzed intelligence data, software and other data are typically transferred as complete files over tactical and strategic network systems. DF Raptor allows lost packets to be recovered without requiring retransmission.

At the sender, each data file is considered to constitute one or more source blocks, and, as shown in Figure 2, each source block is encoded by DF Raptor and transmitted as source and additional repair packets. At the receiver, the successfully received packets are then decoded by DF Raptor to recover the original source block(s).

The number of repair packets generated by the DF Raptor encoder provides fundamental flexibility. A small number of repair packets may not allow recovery of the complete source block in the face of excessive packet loss, while a large number of repair packets will consume additional transmission time that could be used for the support of other network activity. Many different alternative solutions are possible. For example, DF Raptor could be used:

• to continually generate and send repair packets, only ending the transmission upon successful acknowledgment from the receiver(s);

• to generate and send a finite number of repair packets based on the anticipated network environment, then transmit any additional repair packets as may be needed by the receiver(s) as requested through a return channel; and

• to generate and send a finite number of repair packets based on estimates of the network's packet loss, ending the transmission once a specific probability that all receivers have successfully obtained the file is realized.

In sensor networks, video surveillance systems, and other in-theater telemetry applications, data is delivered as a real- or near-real-time data stream. For such streaming applications, DF Raptor can overcome the effects of packet loss independent of the network characteristics, the stream characteristics, or the specific sources of packet loss.

At the sender, the original stream of packets is partitioned into consecutive source blocks of data that are treated independently, as shown in Figure 3. Each successive source block is encoded by DF Raptor, thereby generating more packets than were originally present in the source block. A new stream is then transmitted consisting of the original source packets and their respective repair packets. The overhead introduced by the repair packets directly translates into bandwidth expansion: to maintain the timing of the original stream, the source and repair packets associated with a source block must be transmitted over the same duration as the original data, requiring a faster data rate than the original stream in order to accommodate the additional repair packets.

The number of additional packets can be adjusted depending on bandwidth limitations and on the desired level of packet loss protection. For each source block that has been protected using DF Raptor, as long as the receiver successfully obtains just a few more encoded packets than were in the original data stream, then the original stream can be fully recovered even if many of the encoded packets have been lost during transmission.

At the receiver, the received stream of source and repair packets is processed as blocks and decoded by DF Raptor. The DF Raptor decoder processes successive blocks of the received stream of source and repair packets in turn so that the original source stream is available for playback.

Streaming applications can be optimally designed to maximize transmission reliability and minimize bandwidth overhead, either by employing a constant degree of packet loss protection and bandwidth overhead or by dynamically changing the degree of bandwidth use as the packet loss rate changes. Additionally, because DF Raptor is a systematic code that transmits source and repair packets, playback applications can make use of whatever source data has been received even if severe packet loss precludes DF Raptor's complete recovery of the full source block.

DF Raptor's fundamental capabilities allow a wide range of possible military communications architectures to be addressed.

Bulk data transfers are often part of tactical operations as raw intelligence data is transferred out of theater while processed and analyzed data is returned for in-theater use. Vast amounts of highly perishable data may need to be reliably communicated between U.S.-based centers and tactical or theater commanders.

Over IP networks, transmission control protocol (TCP)-based protocols such as FTP or Telnet are commonly used to support reliable file transfer. With TCP, however, reliability is realized at the expense of bandwidth efficiency. TCP's flow control and congestion avoidance mechanisms respond to long round-trip times, network congestion and packet loss by reducing the window size and the flow of transmitted data. Intercontinental routes will necessarily involve relatively large round-trip times, and, if any packets are dropped, TCP will respond by strictly limiting the transmission rate. In such cases, no matter how much bandwidth may be made available, TCP will limit the possible throughput, as shown in Figures 4 and 5.

DF Raptor, by contrast, permits an alternative way to reliably transfer data over an IP network that maximizes throughput and fully uses available bandwidth regardless of network congestion, packet loss or latency. Because DF Raptor guarantees the ultimate recovery of any lost or corrupted packets, the transport protocol does not need to provide reliability, and file transfer can make full use of the end-to-end data rate.

For example, if user datagram protocol (UDP) rather than TCP is employed to deliver DF Raptor-encoded packets, then the inefficiencies of TCP are not incurred. A connectionless protocol like UDP can realize throughput levels that approach the end-to-end bandwidth because it neither consumes bandwidth with acknowledgments and retransmissions nor restricts the amount of data that can be transmitted.

UDP, however, does not provide reliability and will allow individual packets to be lost. But if the packets are DF Raptor-encoded, DF Raptor will supply the necessary reliability to ensure that the data is received completely. In Figure 4, this approach with DF Raptor would permit throughput of ˜99.5% for all end-to-end data rates; in Figure 5, DF Raptor's realized throughput would be approximately equal to unity minus the packet loss rate for all data rates. Congestion control algorithms such as TCP-friendly rate control (TFRC), fair layer increase/decrease with dynamic layering (FLID-DL) or Web and equation-based rate control (WEBRC) can also be employed as may be necessary to avoid overwhelming all other traffic.

Point-to-multipoint (broadcast and multicast) modes are essential elements of military communications, allowing data to be efficiently disseminated and shared among many users.

In many situations, moreover, complete and error-free reception of the data by all users is imperative. For example, the delivery or update of navigation data, targeting data, intelligence data, code sets or software may require exceptional efforts to ensure successful delivery. The problem is further exacerbated if the broadcast or multicast channel is strictly one way or if the receivers are available only intermittently and impossible to synchronize or coordinate.

Because DF Raptor is a fountain code, it can generate and continually transmit an unlimited number of repair packets. Each receiver can then recover the original data after receiving an amount of encoded data only slightly greater in size than the original data. It makes no difference to the DF Raptor decoding algorithm when the data is received, in what order it is received or which specific packets are received. If an encoded packet is received without error, it can be used in the process to recover the original data. DF Raptor thus wastes no time waiting to receive the last missing data packet while receiving (and discarding) duplicate packets. As a result, DF Raptor ensures reliable delivery in the shortest possible time. In comparison to the alternative of repeated retransmitting of the source data, using DF Raptor minimizes transmission time and bandwidth.

DF Raptor's fountain characteristic allows the possibility of an innovative mixed or multimode approach to ensure reliable data delivery. Because each DF Raptor output packet is generated independently of any other one, a receiver may collect packets generated from the same data but transmitted by different devices. DF Raptor thus provides a solution to enable true disconnection-tolerant networking.

For example, a mobile receiver in a hostile wireless communications environment could receive the same data transmitted by multiple in-theater points of presence in order to achieve high reliability. Packets received from any transmitter can be used to recover the desired source data, allowing the mobile receiver to readily switch from transmitter to transmitter.

This approach requires only loose coordination among the transmitters/channels — they should be configured so that they are sending unique repair packets in order to avoid potential inefficiencies and, of course, be encoding the same source data. Otherwise, there is no need to synchronize their transmission or encoding activities.

Alternatively, the same concept could be applied using multiple channels or multiple links — rather than multiple transmitters that are geographically dispersed; multiple channels or links from the same point can be used to transmit the data to a multichannel or multilink capable receiver. In this case, packets received on the different channels/links can be aggregated to recover the source data, thereby speeding the overall time of data delivery.

Because DF Raptor provides extraordinary levels of packet loss protection with a flexibly controlled amount of overhead, it is ideal for use in military communications. Today, the low processing requirements of DF Raptor allow it to be implemented as application-layer software hosted by general-purpose processors with minimal changes to existing network infrastructure and devices. In the future, DF Raptor will likely become more integrated into strategic and tactical network architectures, supported by communications equipment and protocols at lower layers. As that happens, file delivery and streaming applications will be able to transparently take full advantage of DF Raptor's packet loss protection in strategic and tactical networks.

Alan Jacobsen is director of marketing at Digital Fountain, developers of advanced FEC technology to enhance the quality of communications over data networks. With more than 20 years of telecommunications marketing and engineering experience, Jacobsen holds an MBA degree in Finance from The Wharton School of the University of Pennsylvania as well as a M. Eng. degree in Electrical Engineering and a B.Sc. degree in Physics, from McGill University.

RSS    Save to Del.icio.us  Digg This