Multi-layer Satellite Network Architecture
In this month’s article we will explore the necessity and efficacy of multi-layer satellite networks. Satellite communications traditionally takes a long time to evolve, partly because of the long development time necessary for space hardware with its special ultra-reliable pedigree, close to 7-10 years of service lifetime, and huge financial investment upfront. With respect to the network revolution of the 1980s to the 2000s, space communications lack the corresponding terrestrial network development by as much as 20 years.
Most of the current satellite systems only provide point to point service such as trunking between large sources and sinks and limited usage for individual (stationary and mobile) user access at modest data rates. Except for low rate audio and data service (using omni-antennas) supported by Iridium and Globalstar in low earth orbit satellites, most satellite communications are provided by geosynchronous orbit satellites with the benefit of only requiring static (usually dish) antennas on the ground. Almost all satellites are transponder1 satellites except for a few specialty defense satellites with on-board processing. In the networking functionality context all the protocols are run on the ground. At low data rates (less than mbps), the typical Internet protocols works, albeit not perfectly. When the data rate is high, specifically when there are many packets in flight (>>100) in a round-trip time, e.g. many geo-systems with 250mS propagation times, TCP does not work efficiently and is typically replaced by a custom protocol, tailored to the long time delay, and proxies (which are application layer processes) are often used to interface with TCP/IP in the terrestrial part of the network. This is the first example of a cross layer design to provide reasonable end-to-end service. For transponder satellites, switching and routing are done on the ground and the typical Layer 3 routers work.
At least two new technologies are driving satellite networking today. The first is solid state phase array antenna and amplifier, and the second is smallsat technology. Gallium nitride (even silicon at lower powers) phase array allow beam forming and nulling via antenna processing and also satellite position tracking without mechanical steering antennas for non-stationary satellites. Smallsat technology, including space laser crosslinks, opens up the possibility of low cost low earth orbit satellites with high data rates and low latencies using directional antennas.
As an architecture building block, phase array antennas open up many more architecture options than electronically steerable antenna for no mechanical movement tracking. On the downlink and uplink, multi-beam patterns can be dynamically and adaptively configured to meet non-uniform traffic demands, while massively re-using the same frequency assignments for spatially orthogonal user regions which have big implication for revenue generation with limited expansive frequency leases. To make full use of this new architecture building block in the physical layer, the architecture must involve, in addition to the physical layer, the MAC (media access control) Layer to de-conflict contention by using time multiplexing when more than one requests are present simultaneously in a single geolocation where the users are so close that nulling is not possible. When there are time critical services such as interactive video conferencing, the protocol must have some knowledge of the time deadline of the sessions and inform the MAC Layer and the Routing Layer (as in which of the two satellites in view are free or has the shortest path to route the message to the users or receive messages from the users). It is clear that without multi-layer coordination the above architecture feature is not feasible. That does not automatically say the designer has to entertain such an option, but the imperative is revenue generation using massive frequency re-use and also dynamic assignment to meet the demands of more users with time deadline service requests, albeit with more protocol complexity and thus must be considered in the design trade-off.
A key technology that has never made onto a commercial satellite, though it has been used in the defense sector for several decades, is on-board processing. In the physical layer on-board processing can correct errors on the uplink and power balance the downlink among user sessions for more efficient use of precious downlink power. However, in the past on-board processing electronics are expensive and heavy and power consuming and thus not affordable (high SWAP, Space, Weight and Power). The recent development of CMOS substantially lowers the cost, weight and power consumption. Redundant electronics can be flown without resorting to power hungry and low performance radiation hard electronics. With on-board processing it opens up the possibility of on-board routing by a space-borne router, the key feature that can make crosslink an effective architecture option. Terrestrial routers have high SWAP and to prevent congestion, due to contention, routers are typically balanced at or lower than 20 percent utilization. With the need for low SWAP for space-borne electronics, these levels of loading of the routers is unacceptable and unaffordable. A new generation of routers working at high loads must be created. With routers in space, especially for low earth orbit constellations, routing can be a lot more efficient, lowering both cost and latencies.
Finally, expensive satellite systems need to be highly loaded to recover investment costs, and differentiated service can be used to serve higher paying customers first to optimize revenue. This requires multi-class service detectable by the spacecraft and in conjunction with the MAC, routing and the application layers, optimizing revenue and QoS. To field a financially competitive system will involve differentiated service offering, revenue projection, marketing, operation cost projections and reductions, financing and lifecycle costs of systems. It is the ultimate multi-layer design problem. The final cost to users and expected profit/risks are very sensitive to the best possible design, and the space of all possible designs should be conscious to the architect. There is nothing sacrosanct in the current Internet protocol and architecture that is not within the realm for change.
1 In a transponder satellite, the modulated RF uplink signal is not demodulated on-board the space-craft but simply filtered, perhaps shifted in carrier frequency, amplified and sent to the downlink antenna for transmission.