Читать книгу Dynamic Spectrum Access Decisions - George F. Elmasry - Страница 70

In the previous section, we saw how a distributed DSA engine can interact with a distributed routing engine in order to create a low spectrum footprint MANET with antijamming, LPI/LPD, and dynamic spectrum reuse capabilities. In this section, we will show how this distributed cooperative routing engine concept, which is local to a single network, can be adapted to global heterogeneous routing, which is between the different types of hierarchical MANETs.⁸ The goal here is to explore how far a system that is designed as a hybrid between local and distributed cooperative DSA decisions can go before we move to cases where a centralized DSA arbitrator is needed.

Making distributed DSA work in a global manner with heterogeneous hierarchical networks can be based on the following principals:

 The creation of multitiered DSA engine architecture where a platform that is a node of more than one network (gateway node) has a distributed DSA engine for each network and a parent DSA engine (for global arbitration of DSA decisions).

 The creation of two distinct types of cognitive routing engines where one type is for routing local to the network and the other type is for gateway nodes that can create global routes.

For this multitiered architecture, let us refer to the lower tier DSA engine as the waveform DSA engine, the routing engine local to the network as the waveform routing engine, the upper tier DSA engine as the master DSA engine, and the upper tier routing engine as the master routing engine. Figure 4.6 illustrates how a node that is not a gateway node would contain only the two waveform engines while a gateway node would contain all four types of engines with the lower tier engines (waveform DSA engine and waveform routing engine) having one separate engine for each waveform type the gateway node has. The master DSA engine and the master routing engine will always have a single instantiation in each gateway node.

Figure 4.6 Local and gateway nodes cognitive engines.

Notice that there are different approaches used to create dynamic heterogeneous global routing. One known approach creates an interface control definition (ICD) between each waveform type and the networking layer above the waveform. This interface is sometimes referred to as the router‐to‐radios (R2R) interface. In Figure 4.6, this interface may be replaced with the interface between the waveform routing engine and the master routing engine. The approach covered here relies on the master routing engine making routing decisions by choosing between the different available paths based on the condition of each wireless network in the path. This interface can override R2R protocols such as the point‐to‐point protocol over Ethernet (PPPOE), which have been shown to be insufficient in dynamic MANETs. The approach covered here leaves certain dynamic spectrum management local to each network through the waveform DSA engine and uses the master DSA engine for other DSA decisions. Here, we have a more comprehensive approach than that of Section 4.3 where the gateway node is part of the spectrum allocation negotiation between the different MANETs.

There are established methodologies that address getting statistics that convey the local MANET dynamics from the waveform (radio modem) to the routing layer above. These established methods assume that the wireless links below the routing layer have already been established. Protocols such as the dynamic link exchange protocol (DLEP) communicate metrics to include link quality, bandwidth, and neighbor discovery. With DLEP, the wireless links and communications with neighbors must be established before these metrics can be passed to a routing engine. As such, the master routing engine in the construct in Figure 4.6 will not be able to make routing decisions that include an underlying waveform network unless radio links are established. The multitiered approach illustrated in Figure 4.6 can generalize the DLEP a step further by having the master DSA engine (higher tier engine) perform proactive sharing of the node's links capabilities over a “control plane” medium even when some of the available wireless links are not in use (i.e., in a large scale heterogeneous network, some MANET links have not been formed yet). In the gateway node, the master routing engine is able to receive (via the master DSA engine) location, status, capabilities, and spectrum resources utilization information from peer remote radios engines before all link establishment occurs and use this information to establish potential routes. The master DSA engine is able to use this information to perform spectrum resources deconfliction, link closure estimates, and bandwidth estimation before these links are established. The DSA master engine's ability to perform proactive sharing over different waveforms (mediums), through the master routing engine already established routes, gives the master routing engine the ability to ascertain potential data rates to neighbors on each network type prior to the establishment of all wireless links to these neighbors.

Note that with this type of heterogeneous waveforms formation, the interface between the waveform DSA engine and the master DSA engine has to use unified wireless link metrics regardless of the waveform type which the waveform DSA engine represents. One waveform type may express the health of its wireless links in a different way from another waveform type. This multitiered architecture requires some normalization of the link health metrics calculated by the different waveforms in order for a routing engine to create an optimum global routing table without being skewed to use one waveform over another due to the lack of uniformity of link health metrics. This normalization should also allow the different networks to use different routing approaches internal to the formed MANET independent of global routes. One of the most important essences of creating true seamless efficient heterogeneous networks is to allow each network to use the best protocols for its internal routing and link health representation while requiring the different types of waveform DSA engines to adhere to a unified representation of the network metrics when it comes to global aspects.

Let us illustrate how this hierarchical architecture can work with the case of a low bandwidth waveform that has global connectivity over a specific deployment. This low bandwidth waveform can be used to create a large RF footprint MANET that is used to establish the control plane over the deployed large‐scale heterogeneous hierarchical MANETs. In addition to the low bandwidth global network, the theater deployment has different types of higher bandwidth waveforms that will establish different types of networks (e.g., mesh networks, omnidirectional multiple access networks, LPI/LPD directional networks, etc.) where spectrum resources can be allocated dynamically within these networks and utilized globally by the master routing engine. Let us refer to these higher bandwidth networks links as offload links. The goal of DSA in this construct is to use these offload links dynamically and on‐demand considering the following:

1 When data traffic to a node over the low bandwidth global network exceeds a defined threshold, the master routing engine will ask the master DSA engine to allocate resources over a specific higher bandwidth network. The master DSA engine will allocate the required spectrum resources by asking the corresponding waveform DSA engine to create a flow9 to that node. This action will create a new routing path for the master routing engine.

2 When data traffic levels exceed the currently allocated bandwidth, the master routing engine will ask the master DSA engine to increment the allocated resources. The master DSA engine will ask the corresponding waveform DSA engine to acquire more bandwidth over the established offload data link. The current waveform DSA agent might adjust allocations or the master DSA engine might switch to another offload link on another network as necessary to meet the traffic volume needs. In either case, the master DSA engine will inform the master routing engine if the created routing path has increased bandwidth or if a new routing path is created with the required bandwidth.

3 If traffic levels go below the currently allocated levels, bandwidth over the offload datalinks will be reduced, releasing some of the spectrum resources. Corresponding message flow to the two cases above will ensure the release of resources is known to all involved engines.

4 If traffic levels go below a threshold, the flow over the offload data link will be torn down, releasing all resources back to the offload network.

Notice the following:

 The low bandwidth global network is a lifeline to all nodes and is the start of the control plan. The control plane may use dynamic resource allocation (offload links) if needed.

 All offload networks spectrum become shared spectrum resources pools for allocating spectrum resources per traffic demand.

 The master routing engine is independent from the master DSA engine and the master routing engine has no say on how spectrum resources are to be allocated. This allows the inclusion of heterogeneous waveforms that use different resources allocation protocols within this heterogeneous hierarchical architecture. Each waveform can have a specific waveform DSA agent that works independently of other waveforms' DSA engines. One key aspect here is normalization of waveform metrics such that the master DSA agent uses unified resource allocation terms (e.g., flow) and the routing engines sees created routing paths in terms of unified data rate units (e.g., flow) and source destination pair. For the routing engines, a flow may be associated with a routing cost to allow the routing engine to ascertain the best route (path of multiple over‐the‐air hops).

The use of such dynamic link establishment, link adaptation, and link teardown means the use of a dynamic reactive routing protocol creating route tables that get modified quickly. When there is a steady‐state period, the cognitive engines will be monitoring the allocated flow's health and exchanging information. Information exchange during steady‐state periods can include spectrum health awareness and mobile node geolocations. The shared information will allow the waveform DSA and routing engines to anticipate the need for changes¹⁰ local to the network and react in a timely manner. Shared information will also allow the master DSA and routing engines to anticipate the need for global changes and react in a timely manner.

The master DSA engine is tightly coupled with the master routing engine in different ways. Consider the case when the master DSA engine is sharing spectrum awareness with its peer master DSA agents. The master DSA engine will rely on the dynamically created routes by the master routing engine and control traffic can become traffic demand on its own merit. If the global low‐bandwidth network cannot accommodate all control traffic volume, the offload network's spectrum resources would be used for control traffic.

The available resources on an established data link can change due to mobility, link stability, and link quality. For the cases where the waveform DSA engine learns new information about the link resources, the waveform DSA engine can inform the master DSA engine about any changes. One case of dynamically increasing link bandwidth due to mobility would be when a waveform switches to a lower forward error correction (FEC) mode, increasing throughput at close range while using the same amount of spectrum resources.¹¹ On the other hand, a case of decreased bandwidth is when the waveform must switch to a higher FEC mode due to increased range or jamming. With dynamic MANET, situational awareness must be shared between all cognitive engines, which can result in constant change of routing tables, constant changes in spectrum resource allocation and teardown, and consequently increase in control traffic load.¹²

Let us summarize the advantages of this heterogeneous distributed cooperative DSA approach:

 Having a data plane that is independent of establishing offload links means the ability to exchange spectrum resources and link information before all networks are established.

 Waveform engines can be developed independent of each other as long as the exchange of waveform information is normalized between all types of waveforms.

 Waveforms can use their internal protocols for position awareness, antenna directionality, spectrum resources establishment, and teardown. Waveforms can decide how reactive routes local to the waveform network are established and torn down.

 A large‐scale set of heterogeneous networks can be deployed together and these networks will morph to the unique scenario needs while the waveform cognitive engines manage the configuration of links internal to the waveforms network and the master engines manager global routes.

 Making global dynamic route tables independent of route tables local to a network makes it possible for heterogeneous networks to be connected and optimized seamlessly.

There are also disadvantages of this heterogeneous distributed cooperative DSA approach:

 The amount of DSA control traffic can grow exponentially as the number of nodes in a network increases and as the number of heterogeneous networks increases.

 Some nodes can go out‐of‐synch or temporarily lose connectivity and some information exchange control traffic can be lost. Retransmission of control traffic can further cause an already large volume of DSA control traffic to grow larger and can delay the adaptive route creation decisions.

 There are limits to global views of a large‐scale set of heterogeneous networks through distributed protocols. The master DSA agent in each gateway node will create a spectrum map that is less accurate than the spectrum map that can be created by a centralized arbitrator.

To make different types of IP networks work seamlessly, Internet routing protocols faced similar challenges to what is described here for DSA challenges. This led to the creation of routing areas and gateway protocols such as the border gateway protocol (BGP). In some IP network design, certain routing configurations had to be left for a centralized network manager to decide. Similarly, the reliance on a distributed cooperative DSA for large‐scale heterogeneous networks will always show some limitations and lead us to consider the need for a centralized spectrum arbitrator to curb control traffic volume and arbitrate fairly between heterogeneous networks in certain cases while allowing a human (network manager) to monitor the cognitive network's health and set policies and rule sets. The network manager can change policies when needed.

One simple way to curb control traffic volume is to reduce its rate. Figure 4.7 illustrates a conceptual view of how control traffic volume can depend on the updating intervals. As updating interval periods decrease (more frequent exchange of DSA messages occur), control traffic volume increases. On the other hand, as the updating interval periods increase (less frequent exchange of messages occur), control traffic volume decreases. There is a desired cutoff limit of the amount of DSA control traffic depending on the heterogeneous systems under design. The limitation of the amount of DSA control traffic can lead to curbing message exchange rate and the impact of this limitation can make resource allocation and dynamic routing tables less dynamic and less reactive to the deployment needs. This can be harmful, especially for fast‐moving nodes in very dynamic MANETs.

Figure 4.7 Conceptual view of how control traffic volume depends on updating rate.

Notice that within the heterogeneous MANET designs presented in this section, the distributed cooperative DSA¹³ has a mix of local DSA decisions made by the waveform DSA engine and distributed DSA decisions made between the master DSA agents. Although these systems are referred to as distributed cooperative DSA systems, in essence they are hybrid between local DSA and distributed DSA decision systems with two different hierarchies of distributed cooperative decision fusion. Similarly, adding a centralized arbitrator to this construct creates a different type of hybrid DSA design that has more hierarchical DSA entities.