One of the first technologies used in communications satellites to deliver Internet connectivity was SCPC (Single Channel Per Carrier) which refers to using a single signal at a given frequency and bandwidth. In an SCPC system, satellite bandwidth is dedicated to a single source, in other words you buy your own part of the satellite. The primary advantage of SCPC is that the architecture allows the complete direct and synchronous connectivity between the two remote points on Earth (the remote customer and Internet connectivity teleport for example), that's why it is now widely used to provide dedicated backbone links between Global Internet access and Telephony Service Providers worldwide.
The step forward to a consumer market was an introduction of shared services like Frame Relay. Frame Relay circuits enable Internet Service Providers to create a shared bandwidth resource: it was designed to make more efficient use of existing physical resources, thereby allowing the overprovisioning of data services to the customers, as most of them were Home/SOHO/SME users and Internet cafes, they are unlikely to be utilising a data service 100 percent of time.
Just as the Frame Relay, a number of satellite Internet access systems built on the TDMA architecture have conquered the Home/SOHO/SME, Hot Spot and Internet cafe market. The burstable nature of an IP traffic, combined with the technical possibility of shared resource solutions like Frame Relay or ATM to provide several Private Virtual Circuits (PVCs) inside the single physical resource, has offered the opportunity for Internet Service Providers to offer a low-cost shared broadband Internet access service by overbooking the physical resource, as mentioned above. A term "Contention Ratio" has been introduced, which is also known as Sharing Ratio or Oversubscription Ratio.
There is a wide range of shared satellite Internet access services available on the market nowadays, built on different TDMA technologies (TDMA, MF-TDMA, D-TDMA, ...), contention ratios (from 1:50 to 1:2) and prices (from few tens to several thousands per month) without any clear buying guidelines for the user to understand which service is exactly applicable to his tasks.
The first thing to understand is the concept of sharing or subscription ratios. The fundamental idea of oversubscription is that as ISP provides bandwidth for users, the total amount of bandwidth assigned exceeds the actual port speed that is being delivered. The idea is to make sure that "network" bandwidth is available when users need it, and that the access point does not become a bottleneck that creates a bad user experience.
As mentioned above, the fundamental reason for oversubscribing is to safely conserve expensive bandwidth and utilize it to it's maximum potential. Bandwidth that is unused and idle, is wasted, and it becomes a cost which is not producing revenue for the Internet cafe or Hot Spot operator or simply forms an enormous service price which individuals can not afford. On the other hand, if bandwidth is far oversubscribed and unavailable when needed by users, then customers will be unhappy with performance and will take their business elsewhere. So a balance must be struck, to maximize revenues and profits, while maintaining and growing a satisfied customer base.
The intermittent, bursty nature of Internet traffic virtually guarantees that all users are not going to require maximum bandwidth at the same time. This is similar in concept to the telephone company, which has been playing this same type of game for years by oversubscribing users to the number of phone switch ports available. The phone company knows that not all subscribers are going to be actively making phone calls at the same time. They are able to share or oversubscribe the switch ports based on the historical call volumes made by subscribers. When too many people try to place a call at the same time, a busy signal is generated. If this happens too often, subscribers become upset with the service provider.
Oversubscription is deeply rooted in the fundamentals of statistical multiplexing. Because data traffic tends to be quite intermittent and bursty, statistical multiplexing is able to save money by optimizing the use of the available bandwidth. Time division multiplexing (TDM) technology permanently assigns each individual channel a portion of the bandwidth. If a channel is idle, the bandwidth is wasted and gone forever.
Taking a look at a simple example, think back to the good old days before PCs, when asynchronous terminals were connected to mini-computers. If you had eight terminals at 1200 bit/sec connecting to the computer with time division multiplexing, they would share a 9600 bit/sec circuit (8 x 1200 = 9600). With statistical multiplexing, the multiplexer has the intelligence to assign bandwidth as needed. For example, in the above situation, one might set each of the terminals to 4800 bit/sec, resulting in a four-to-one (4:1) oversubscription (8 x 4800 = 38,400 and 38,400/9600 = 4). The odds that each of the 8 terminals would be active simultaneously is small enough that one would typically not see a performance degradation. In the event that all of the 8 terminals were active at the same time, the buffers in the multiplexer would temporarily store the data until bandwidth became available again.
This same type of oversubscription takes place in regards to traditional terrestrial Internet access services. ISPs "play the odds" using traffic engineering techniques so that they can provide services to large numbers of customers using shared, rather than dedicated bandwidth. The challenge for network architects is to determine how much bandwidth they need in order to optimize the use of this expensive resource, while providing a service that will ensure satisfied customers. This is a challenge in the terrestrial or wireline industry, just as it is when using broadband satellite for backhaul service.
There are basically two types of wireline services that are generally used for Internet backhaul service:
1. Private T-1 or FT-1 (fractional T-1). These are dedicated point-to-point links used by a single client, often to connect a remote office location to a headquarters location, or to a POP or Point of Presence on the Internet backbone. This type of circuit is frequently used in private networks, but is seldom provided for Internet access.
2. Internet T-1s. These are T-1 circuits whose bandwidth is shared by multiple clients using statistical multiplexing techniques. This is accomplished by aggregating or combining multiple Internet T-1s and then connecting them to the Internet POP, such that the total bandwidth of all the T-1s exceeds the size of the connection to the POP. For example, if no sharing or oversubscription takes place, 28 T-1 (1.54 Mbps) circuits could be connected to a single DS-3 (44.7 Mbps) link to the POP. However, service providers will generally oversubscribe Internet T-1s with 1:8 to 1:12 ratios, so that anywhere from 200 to 300 T-1s might share bandwidth on a single DS-3 circuit.
Many new Internet cafe, WISPs (wireless ISPs), and Hot Spot operators are unaware that the Internet T-1 they use for Internet access is already oversubscribed, and that this has an effect on the amount of bandwidth that they can deliver to their clients. Most service providers are unwilling to discuss their oversubscription ratios with their customers, so it pays to do some research and talk to other customers of the carrier in order to get some idea of what kind of service to anticipate.
It is not necessary to be alarmed or overly concerned about this practice. At the very high speeds that service providers are aggregating Internet T-1 circuits, the statistical multiplexing is very efficient and bandwidth may be shared by many clients quite successfully. Generally the relative price of the circuit compared with other carriers, is a good indication of the subscription ratio.
Similar considerations must be acknowledged when using broadband satellite for backhaul. Dedicated bandwidth, (generally SCPC) similar to Private T-1 circuits, will deliver the entire bandwidth to a single client. Often times, this may be a highly inefficient use of the bandwidth as the bursty, intermittent nature of TCP/IP traffic means that much of the available bandwidth may be idle and wasted.
It may be more cost-effective to use shared bandwidth as a backhaul service, since dedicated SCPC bandwidth is very expensive. An Internet Caf? or Hot Spot (ISP) operator may only be able to afford a low-speed SCPC circuit, for example 64 Kbps x 32 Kbps. Using shared bandwidth, he might be able to lease a higher speed backhaul circuit, for example 512 Kbps x 64 Kbps for the same cost. This allows his users to receive faster downloads, and to transmit at faster speeds.
He should understand however, that if the broadband satellite network is very busy and the Satellite Network Operator (SNO) has oversubscribed it too much, that he could temporarily end up during peak traffic times with less bandwidth than a lower speed SCPC circuit would provide. So far this is the most often admitted mistake in the satellite broadband market, especially with the entry-level TDMA and MF-TDMA VSAT systems when users, mislead by the lack of purchasing guidelines, buy highly oversubscribed services and experience a poor quality of Internet connection.