Skip to main content
Publications lead hero image abstract pattern

Publications

Multi-Layer Wireless Network Architecture

In this month’s article we explore the necessity and efficacy of multi-layer wireless networks. To understand the imperative for a coordinated multi-layer architecture, we should examine the types of future applications that will use the infrastructure. Future wireless networks will do more than just provide an increase in data rates on traditional applications. They will provide real time analytics upon request or pushed, geo-location for many applications including some needing critical time-deadline services such as vehicular safety, disaster warning and relief, clique interactions in virtual groups of common interests, and other smart city applications, etc.

Some major network attributes (by no means an exhaustive list) driven by applications that have come to the forefront are:

  1. Increase of data rate for applications up to 2-3 orders of magnitude.
  2. Large transactions (usually in the form of big flows) that result in dynamic and bursty non-uniform spatial traffic demands. Surges can be geographically concentrated and correlated, preventing the network from being optimized for average spatial performance. Hence, congestions cannot be mitigated by fair throttling or slow load balancing.
  3. Delay-tolerant but highly reliable data transactions.
  4. Delay-sensitive communications; latency outweighing throughput.
  5. Interconnections of, and provide services in junction with, centralized metro cloud computing centers and edge computing (FOG).
  6. Policy driven services with premium class requirements.
  7. Emergency/first-responder and disaster relief services.
  8. High assurance and secure/resilient network infrastructure and services.

The following are some technology building blocks that are in play that will affect achievable network architectures:

  1. Multiple antenna and spatial processing by data center or edge computing resulting in massive frequency reuse via nulling for up and down links.
  2. Agile spectrum access from GHz to THz.
  3. Congestion control and performance optimization via Media Access Control (MAC) of access points and backhaul fiber network and computing load balancing.
  4. Backhaul agile all-optical fiber lightpath switching (as fast as time scales of ∼10 mS) and optical and THz wireless with fast topology and capacity reconfigurations.
  5. Massive FOG edging computing and metro data centers.
  6. ML/AI algorithms for network management and control, computing resource allocation and network security/resilience.
  7. GPU, programmable gate-arrays and other specialized processors for massive stream processing at network nodes.
  8. Cognitive network management and control algorithms including statistical techniques, MI/AI and math programming.

To realize many of the network attributes given above, the network architecture needs to work across multiple layers including computing resources and applications, and in a way that optimizes performance and costs. A smart city network is a subset of a metro wireless/fiber network. Generally, this network must have certain features or architecture constructs to make the infrastructure perform the span of diversified heterogeneous services. Taken together, these architectural constructs form a coherent basis for networking in the Smart City/metro networks of the future. All these are implemented via a centralized or distributed (harder to design) network orchestration engine. For practical implementation of this network, designers must consider the costs of various architectures including life-cycle costs and financing. Some of the major features of the construct of this network include but are not limited to:

  1. MAN infrastructure:
    1. Dedicated Metro/Smart-City data center/s: Attaching to the data center should be a network management entity connecting local area Smart City networks to the existing metropolitan area networking infrastructure.
    2. FOG edge computing for low-latency and low-cost applications: Edge computing is enabled to make decisions and instantiate applications independently of the centralized cloud for low latency and responsiveness.
    3. Dynamic and agile topology reconfigurations via optical bypass in fiber and optical/microwave wireless links.
    4. Priority service for critical traffic at access points, MAN routers, data center and edge computers.
    5. Both access points and users equipped with fast tunable directional multi-elements antennas; RF signals from multiple access points brought to the data center for collaborative spatial processing.
  2. Network orchestration engine:
    1. Policy driven networking and congestion control via orchestrated multi-layer wireless/fiber optimization, vectoring service request to the edge or the data center: Via orchestration, the link costs and/or path weights in the routing tables of edge routers can be quickly updated to bias traffic flows toward less congested network paths.
    2. Load-balancing at the edge via physical and virtual topology changes: If upstream traffic is overloading the network, the network orchestration engine may “drop or reduce the quality of service” of certain applications to FOG servers adjacent to edge routers in the MAN.
    3. Priority wireless access reservation for critical traffic: configuring MAN routers to provide (pre-emptive and non-preemptive) priority service to critical traffic at queues.
    4. Dynamic access point assignment in coordination with the backhaul network and computing/storage services: MAC protocol to guarantee network access for critical users within time delivery limits predicted by the network orchestration engine such that it can pre-empt non-critical users by reassigning them to other access points or wait in the queue; the access point to computing device assignments must be close to the edge to avoid propagating large amounts of control overhead through the network.
    5. Agile resource augmentation: Additional resources such as UAVs or terrestrial robots can be deployed as alternative communications media to supplement existing overloaded access point capacity.
    6. Reducing delay by striking a balance between computational speeds at the edge computing servers and transport speeds and congestion states in the MAN.

From the construct of the network it is clear that an optimized multi-layer network architecture is necessary. Thus, the traditional task of congestion control (mostly a Layer 3 function) will, in the future, involve, antenna beam forming and nulling in both the up and down links, access point assignment via MAC, routing in the backhaul in both the Physical and Routing Layers, and computing resource assignment and applications coordination.

There has been an explosion of network technology building blocks in recent years. The situation is not unlike the example of the changes in building architectures at the turn of the century. Whereas previous buildings (typically several storeys high) are made from wood, bricks and glass, the introduction of steel and reinforced concrete has opened up the new architecture space of doomed-stadiums, large shopping malls, sky-scrapers, etc. In fact, with these new technologies the “sky” is the limit (pardon the pun)! There is no reason why the researcher should be compartmentalized and confine their pursuits to a thin layer of the network.