Wireless dynamic spectrum access (DSA) networks require: (1) knowledge of available spectrum through wide-band spectrum sensing, policies, etc. [1]; (2) real-time spectrum management–viz., provisioning and release of RF bandwidth; and (3) a network infrastructure and/or endpoints that support these technologies. The focus of this paper is requirement 2.
Wireless DSA networks may require “new architectures and associated signaling and control protocols” for their real-time spectrum management component [2]. However, wireless DSA networks, and optical networks that support real-time dynamic provisioning of wavelengths, have very similar spectrum management requirements. Hence, we believe: (1) existing optical control plane architectures and protocols that support real time spectrum management can be adapted to manage wireless spectrum; and (2) methods used to improve the performance of dynamically provisioned optical networks can also be applied to commercial-sector wireless DSA networks.
We review basic spectrum management requirements for commercial-sector DSA networks in Section II, briefly describe a candidate control plane architecture, protocols, and performance issues in Sections III and IV, describe usage scenarios in Section V, and our conclusions in Section VI.
SECTION II
BASIC REQUIREMENTS
Wireless DSA networks deployed in the commercial sector will require a number of real-time spectrum management functions; several are especially challenging:
- Ultra fast provisioning and release of RF bandwidth to make efficient use of limited spectrum, to operate in environments that may change many times per second, and to take advantage of short-lived RF micro-environments;
- Support for endpoints separated by multiple hops, and by multiple administrative domains;
- Support for unicast, multicast, and broadcast;
- Support for a broad range of wireless DSA usage models; e.g., where users may hold exclusive rights to their assigned spectrum; where users may ‘lease’ spectrum in (near) real time and hold it for seconds to days; where users may preempt and seize spectrum in real time and hold it for seconds to days (e.g., emergency services); where users may share licensed spectrum by varying their usage in space or time; where users may operate over blocks of pre-provisioned but unassigned spectrum; where users may operate over shared unlicensed spectrum; and where ‘private commons’ licensees may sublease spectrum to users that do not fit within traditional dynamic spectrum leasing arrangements (e.g., peer-to-peer devices that do not use the licensee's network infrastructure).
SECTION III
CANDIDATE ARCHITECTURE AND PROTOCOLS
A number of optical networks use wavelength division multiplexing technologies. A wavelength is a band of optical spectrum that may be assigned–statically or dynamically–to a single user or source, or used to carry aggregate traffic that has been multiplexed in some way. And it may be provisioned in (near) real time, and held for milliseconds to months.
A. “Just in Time”
JIT (“Just in Time”) is a family of simple, fast, out-of-band signaling and control protocols used to dynamically provision and manage wavelengths in optical networks [3]. JIT is independent of media type, and thus can also be used as a broker of wireless spectrum. JIT is an open, published architecture, and its protocols have been implemented in hardware. Software-based implementations are also available.
JIT has been deployed and tested in the Advanced Technology Demonstration Network testbed in Washington DC [4], which is part of the us Defense Department's Global Information Grid Evaluation Facility, and which serves a number of science, technology, and engineering programs [5].
B. Reference Architecture
A reference architecture for wireless DSA networks assumes that a large number of communications devices compete for various parts of the RF spectrum. Devices have varying power and signal sensing and transmission characteristics.
Devices can use the JIT protocol to create requests for spectrum allocations by sharing a low bandwidth common signaling channel. Devices may use other components of the JIT control plane (e.g., intra- and inter-domain routing), to support additional DSA functions–spectrum brokering, preemption, policy management, etc. We briefly describe some of JIT's features in the remainder of this Section.
1) Signaling Channel
Out-of-band signaling is technology-neutral; i.e., the signaling and data planes are orthogonal, so JIT signaling can be used to support a wide range of DSA usage models, and to dynamically assign wireless spectrum in a number of ways–based on frequency, time, space (location or direction), signal characteristics (power, coding) or combinations of these [6], [7]. Most have direct counterparts in optical networks.
A likely implementation for wireless DSA networks is to use an ultra-wideband (UWB) common signaling channel that does not interfere with other communications, and that may have a relatively low bit rate. Spreading codes can be used over the UWB common channel to emulate point-to-point signaling paths between node pairs if required.
2) Message Structure
JIT control messages (i.e., signaling, routing, network management, etc.) have hardware-parsable components with hop-by-hop significance, and software-parsable components with end-to-end significance. The same message format is used by all of the JIT management protocols, which greatly simplifies their implementation in hardware. The structure allows hardware and software to parse their respective segments in parallel.
3) Provisioning Schemes
JIT can dynamically provision spectrum in real time. Spectrum requests may vary in capability (e.g. SNR, interference footprint, path reliability) and holding time, and may be spread over multiple administrative domain(s).
JIT can suggest an alternate spectrum band if a request cannot be met, and can deterministically delay a request if a specific band is scheduled to become available shortly.
Spectrum may be released in two ways. Implicit releases are timed, and require the holding time to be conveyed with the allocation request. Explicit releases issue a control message.
4) Signaling Speed
DARPA XG spectrum sensors obtain knowledge of available spectrum in tens of microseconds (for a 25–50 MHz sub-band) [1], which suggests that provisioning must be equally fast. JIT hardware is able to generate signaling requests in 13 microseconds (using a 400,000-gate field programmable gate array), and in 1.3 microseconds (using a million-gate FPGA). Signaling request times can be reduced to 725 nanoseconds with higher speed FPGAs, and to about 200 nanoseconds with pipeline processing and parallelization of control message functions. A software implementation using a 1 GHz processor issues requests in about 1 millisecond.
This has two important implications. First, traffic in RF environments is inherently bursty. Coupling fast DARPA XG-class spectrum sensing with ultra fast JIT signaling means that spectrum utilization can be significantly increased by reserving spectrum only for the duration of the burst.
Second, ultra fast signaling means that application- and network protocol-initiated provisioning and release of wireless spectrum is technically feasible. E.g., spectrum can be provisioned and released on demand by applications like HTTP and by transport protocols like TCP, rather than being provisioned and held for an entire session.
5) Policy Distribution
Dynamic spectrum management usually requires devices to monitor utilization, and to request unused or underutilized spectrum as needed. Devices in close proximity tend to have similar spectrum profiles and may make similar choices with respect to channel selection. Using JIT to share information about spectrum availability and to make explicit reservations for spectrum segments can increase spectrum efficiency and support a higher density of devices in the same area.
Endpoint radios may have policy rules that apply at their locations. If so, JIT can be used to distribute policy rules. Multicast is an effective way to distribute policy. JIT supports several different types of multicast joins (source-managed, leaf-initiated) and scope limits [8].
6) Security
JIT's out-of-band signaling protects network functions from unauthorized use by functionally isolating the control channel from the data channels. JIT supports encryption and strong authentication of users prior to joining the network.
SECTION IV
PERFORMANCE ISSUES
A. Blocked Requests
Estimates of how much allocated spectrum is in use at any given moment (on average) range from 2% (overall) to 13% in the heavily used 30–3000 MHz band [9]. If an endpoint has potential access to multiple 5–20 MHz spectrum channels, then the approximate probability of having no spectrum available when a request is made (i.e., the request is blocked) is a function of the number of candidate channels and offered load (Figure 1). The approximate blocking probability is well below 10−4 for any combination of four or more channels and for offered loads up to 50%. Note that even a few channels can support loads up to 30% at blocking probabilities below 10−2.
Figure 1. Approximate blocking probability as a function of the number of candidate channels and offered load.
B. Provisioning Schemes
JIT uses “tell-and-go” (TAG) provisioning in optical networks to greatly reduce the time required to allocate spectrum. TAG does not wait for acknowledgments from endpoints or intermediate devices (e.g., wireless routers) before transmitting. JIT also supports “tell-and-wait” (TAW) provisioning with reliable signaling in which DSA requests are acknowledged prior to use and which incur a full round-trip delay plus processing time. TAG provisioning is most effective when the blocking probability is low (Figure 1), and may be appropriate for some wireless DSA networks.
Figure 2 shows the differences between TAG and TAW signaling for transmissions that transit multiple hops or multiple administrative domains.
A TAG transmission is preceded by a JIT SETUP
message sent over the out-of-band signaling channel. The source does not await confirmation that an end-to-end path has been established; instead, it begins transmitting after receiving a SETUP ACK
from the first device in the multiple-hop path. The SETUP
message informs each device along the path of the impending data transmission so that spectrum can be allocated prior to the transmission's arrival. Spectrum may be implicitly released if the holding time is known and conveyed with SETUP
, or explicitly RELEASEd
as in Figure 2. JIT supports preemption, so a transmission may be allowed to preempt a lower priority spectrum allocation.
Figure 2. TAW spectrum provisioning, transmission, and explicit release over a multiple-hop path. Times are not to scale.
A TAW transmission is also preceded by an out-of-band JIT SETUP
message (Figure 2). However, the source does not begin transmitting until it receives confirmation that an end-to-end path has been established. Spectrum may be either implicitly or explicitly RELEASEd
, and lower priority assignments may also be preempted.
TAG provisioning is important in optical networks where error rates are negligible, and where blocking probabilities are relatively low at offered loads up to 50%. TAW provisioning adds unacceptable delay overhead in these networks–e.g., a 5 MByte burst that can be transmitted in a millisecond over a 40 Gbit/second optical channel may incur several seconds of TAW signaling overhead.
TAG provisioning is less important in wireless networks where signaling channel error rates are relatively high and data channel bit rates are relatively low. TAG's advantage in wireless spectrum efficiency–i.e., transmission time ÷ total holding time1, may be negligible at holding times above 10 seconds if fast provisioning is available. Confirmation of end-to-end paths provided by TAW signaling is also important in lossy wireless environments. As noted, JIT can support both TAG and TAW provisioning in the same network, so the choice may be left to the application.
SECTION V
USAGE SCENARIOS
JIT signaling can operate transparently with all of the regulatory and market-based models proposed for commercialsector wireless DSA networks–exclusive use, dynamically assigned rights, licensed use with block sub-assignments, secondary leasing, commons (“sense and adapt”), etc. [10], [11].
JIT can also operate transparently with DARPA XG-class “sense and go” opportunistic adaptive cognitive radio architectures in which endpoints (e.g., software-defined radios) use (near) real-time sensing and JIT signaling to match transmissions with idle spectrum. Transmissions in this usage scenario are typically short-lived “bursts”, and may be well-suited for JIT's TAG provisioning2, Opportunistic adaptive systems may also use JIT protocols to share information about spectrum usage with other devices (endpoints, brokers).
JIT signaling is an excellent match for so-called “coordinated DSA networks” in which access to the spectrum in a region or administrative domain is managed by a spectrum broker [2]. Wireless spectrum is ‘leased’ to endpoints, and the lease lifetime may be very short3. This may be a more realistic way of implementing commercial-sector wireless DSA networks. JIT signaling can be coupled with other JIT architectural features (e.g., intra- and inter-domain routing) to effectively manage spectrum in these scenarios [12].
As noted, JIT is sufficiently nimble to support dynamic spectrum provisioning requests from a number of sources–network operators, spectrum brokers, users, applications, and network protocols.
Optical control plane architectures and protocols that support real time spectrum management–provisioning and release of optical bandwidth–can almost certainly be adapted to manage RF spectrum in general, and commercial-sector wireless DSA networks in particular [13], [14].
JIT is an enabling technology for spectrum management in these networks. Its open architecture and protocols are well-suited for managing devices and entities that compete for wireless spectrum.
JIT supports dynamic spectrum provisioning and release, distribution of policy to devices and endpoints, and operates in environments that may change many times per second. Unlike some architectures, JIT also supports multicast, endpoints separated by multiple hops and/or multiple administrative domains, and a broad range of usage scenarios.
Optical testbed trials have confirmed that a hardware-based JIT implementation can provision spectrum with signaling times on the order of a microsecond, and can manage spectrum with holding times ranging from a few milliseconds to months.
Acknowledgment
The authors acknowledge the valuable comments and suggestions of Ed Thomas (Chief, Office of Engineering and Technology, US Federal Communications Commission), Bob Lucky (Chair, Technological Advisory Council, US Federal Communications Commission), Bob Schneider (Joint Spectrum Center, US Defense Information Systems Agency), and Hank Dardy (Chief Scientist, Center for Computational Science, US Naval Research Laboratory).