Home | 3G | Satellite | GPRS | TCP IP | HTML    

IP Transport over Satellite
Satellite systems have long been used to provide telephony, video, private network and data communications. Explosive growth of broadband data services worldwide has renewed interest in satellite systems for the transport of Internet Protocol (IP)-based data.
For long, the satellite communications industry has been constrained by the legacy of serving telephony market. Until now, almost all the available hardware was designed to transport circuit switched voice traffic over satellite. The Internet Protocol Processor Platform (IP3) offered by Kromos Communications, Inc — is the first solution designed for transporting IP traffic via satellite. The features of IP3 were designed to maximize effectiveness in a wide range of applications. Internet service providers (ISPs), network backbone operators, system integrators, and geographically distributed enterprises benefit greatly from the advanced capabilities of the IP3. We will discuss the unique features and applications of iP3.

The Satellite Advantage

As IP traffic volume increases, so does competitive pressure among Internet service offerings. As a result, service providers continually seek ways to reduce costs, improve functionality, and ensure quality of service (QoS) for Internet- based applications. These issues are easily addressed with satellite links. Customers quickly realized the value of transporting Internet-related traffic via satellite. As a result, the market for satellite-based Internet link capacity quadrupled in 1998 (per Internet Via Satellite 99 Report, http://www.spotbeam.com/mansum99.htm).Distance Insensitive Service.
Internet traffic routed via terrestrial networks is often subject to bandwidth limitations and latency due to congested routers and Internet backbones. IP datagrams queue at congested routers, where they contend for router and backbone capacity. They are then routed based on order in which they arrived in the queue or on other criteria, such as the Type of Service (ToS) required. As distance increases between the source and destination address, so does the number of routing points, connection latency, and the likelihood of congestion.
Alternatively, a geo-synchronous earth orbit (GEO) satellite parked at 22,500 miles altitude can provide simultaneous direct access to large geographic areas— as large as one-third of the earth’s surface, depending on the type of beam used. Satellite exploits beam coverage to deliver IP data in a single hop (with minimal end-to-end routing) to one or more locations within the beam

Multicast Distribution

Terrestrial network topology does not lend itself easily to point-to-multipoint
(i.e., multicast) distribution of IP traffic across the Internet. In terrestrial networks, multicast traffic requires a simultaneous delivery of the same information to a number of hosts along a delivery tree (to individual destinations of a class D address) and substantial replication of datagrams. In addition, routers must periodically track each host’s membership activity in order to avoid unnecessary datagram replication and delivery to inactive hosts that taxes network capacity.

In contrast, IP data can be multicast simultaneously to any number of points within the coverage area with a single transmission, rather than as multiple point- to-point transmissions on terrestrial networks. By delivering IP multicast traffic via satellite, complex routing algorithms and management of delivery trees are virtually eliminated. (Figure 1).

Traffic Asymmetry
Internet traffic between hosts and servers is statistically highly asymmetrical— large file downloads by hosts vs. limited traffic from IP requests or acknowledgements. The approximate asymmetry ratio can be as much as 10:1. However, service providers charge for long-distance terrestrial IP transport via coaxial cable, microwave, or fiber optic links, symmetrically—customers pay for the same amount of bandwidth in both directions, regardless of how much they actually use. Because international IP traffic is heavily outbound from the U.S. where the most content servers reside, the U.S.-bound capacity of symmetrical terrestrial links remains significantly underutilized.

Satellites, on the other hand, adjust easily to asymmetrical traffic. Their bandwidth is equally available to all the traffic within a beam, independent of which direction traffic flows.

Quality of Service
Certain applications (e.g., videoconferencing, streaming audio/video, or voice
over IP) have limited tolerance for the unmanaged, contended-access nature of the terrestrial Internet. IP connections over long distances may result in IP datagrams taking significantly different routes and arriving at the destination out of
sequence, resulting in jitter, outages, and reduced QoS. Reserved routes between the origin and destination for the duration of the connection can solve the problem but will prevent bandwidth from being shared by other applications. A more practical implementation is an end-to-end managed network. A satellite link is a managed end-to-end connection that can deliver datagrams in the sequence provided by the originating earth station terminal.

Quick Deployment
Service providers can accelerate deployment of revenue-generating services by deploying satellite links, instead of working with slow telecommunications carriers to install leased lines. Quick-deployable satellite terminals can be installed and activated in less than an hour; small or medium-sized fixed earth stations can be installed in a matter of days. This makes satellite an ideal solution for communicating with remote field operations or areas hit by natural disaster.

The Internet Protocol
With the rapid growth of Internet traffic, the TCP/IP suite of protocols has
become the dominant data protocol used in terrestrial internetworking. Originally, these protocols were intended to maintain data communication in the aftermath of a nuclear war. Therefore, they were specifically developed to work over unreliable, unmanaged, and partially destroyed networks.

Protocol Layers
TCP/IP connections were designed to work in the lower four layers of the seven- layer ISO Open System Interconnection (OSI) reference model. The four layers are the Physical, Link, Network, and Transport Layers. IP-based applications, such as the Web, e-mail, or network management systems, reside in the upper three layers of the OSI model. These three layers are often treated as a single application layer. A basic understanding of the protocol layers helps to explain how TCP/IP protocols interact in providing a connection. Appendix A describes the layers in greater detail.

The lowest, or Physical Layer, is the medium that transports electrical or optical signals— a cable, a wireless link, or an optical fiber link. The next layer is the Link Layer, which is commonly a LAN, WAN, or serial data line and uses its own protocols depending on the specific medium used. The next higher layer is the Network Layer. IP operates in this layer and transports datagrams across the lower layers to deliver them to the destination. However, IP does not provide a guaranteed delivery of datagrams. For this reason it is called connectionless. TCP/IP was designed to work transparently across virtually any type of Physical or Link Layer media.

For a reliable point-to-point connection, the Internet requires a protocol such as the connection-oriented Transmission Control Protocol (TCP). TCP operates in the fourth layer, the Transport Layer. TCP achieves a reliable connection by sending and tracking all datagrams, re-transmitting lost datagrams, and passing datagrams in the proper order at the destination to the Application. TCP compensates for latency differences between the various paths over which the individual IP datagrams may have traversed the network. To verify proper delivery of all datagrams, TCP sends receipt acknowledgements from the destination back to the source. TCP also has the means to prevent and recover from network congestion. However, all these make a TCP connection susceptible to the roundtrip time (RTT) of the network.

Network Congestion Control
In order to guarantee data delivery, TCP must be able to prevent and recover from network congestion. TCP monitors network throughput capacity and adjusts the data flow accordingly. It uses two algorithms—Slow Start and Congestion Avoidance—to control the amount of data in transit. When TCP detects congestion, it reduces the rate at which new datagrams are sent. The appropriate algorithm is selected to control the process of ramping back up to maximum throughput under the new network conditions.

A connection’s maximum throughput is limited by the supported data rate of a link. However, actual throughput is further limited by TCP’s Congestion Window. TCP increments the value of the window by the size of each datagram sent. The window value is decremented each time the originating address receives a confirmation that a datagram has been delivered. The Slow-Start and

Congestion-Avoidance algorithms control flow through the network by modulating window size according to the transmission acknowledgements generated by TCP.

Terrestrial networks normally work with a high signal-to-noise ratio (SNR) and virtual absence of bit errors. Thus, TCP was not equipped with any special algorithm for datagram corruption due to bit errors