Wireless Metro WiMAX Backhaul : Licensed vs Unlicensed by Erik Boch, CTO & VP of Engineering 8/30/2006

There are typically a number of layers to be considered in any metropolitan network design. In many cases, these networks will often rely on existing metro fiber for the central metro-core layer, building out new [wireless] layers for the access and metro backhaul segments. Within the backhaul layer, perceived spectrum scarcity can make the use of unlicensed technology an attractive alternative to licensed technology. This paper examines the wireless backhaul layer with a view to illuminate desirable performance attributes and how these can be realized with either licensed or unlicensed technologies. Typical Metropolitan WiMax Network Topologies Focusing on wireless networking, there are a number of general topologies. These include: (1) WiMax Multi-point Access + Point-to-Point (or daisy-chained) Backhaul The typical configurations of this topology are illustrated below in Figure 1 and Figure 2. The key difference in these topologies are that daisy-chaining reduces overall path availability performance and increased delay and delay variability. The daisy chain, however, allows the effective reach of the metro fiber PoP to be considerably extended. Figure 1 � WiMax MpT Access + PtP Backhaul (Illustrative subset of a metro network portion shown) Figure 2 � WiMax MpT Access + Daisy-Chained PtP Backhaul (Illustrative subset of a metro network portion shown) (2) WiMax Multi-point Access + Meshed Point-to-Point (PtP) backhaul Figure 3 � WiMax MpT Access + Meshed Backhaul (Illustrative subset of a metro network portion shown) In this topology, an Ethernet mesh is used to aggregate and backhaul WiMax hub sites, delivering the traffic to the metro fibered PoP location (see figure 3). In this case, the network delivers superior path availability due to the inherent angle diversity and location diversity within the meshed backhaul layer. These gains can provide 5 � 10 X improvements in availability. Additionally, compared to the single layer PtP topology of Figure 1, the meshed solution often results reduced [average] path lengths, further enhancing availability performance. As a general goal, a key design objective of the network design is the minimization of delay and delay variability. Delay performance directly impacts the operation of time-sensitive applications. These include things like VoIP and VIDoIP, TDMoIP services. Delay variability affects the operation of hand-off processing applicable to mobile applications. This variability can take several descriptive forms: Packet flows on a given path suffer variations in their delay traversing the network (due primarily to collisions) End-Site to end-site delay variations resulting from a combination in packet-jitter and differential path delay Usually, low delay Ethernet performance network design requires ultra-low delay network elements, careful traffic segregation (normally using VLANs), flow prioritization and cut-through processing. Low delay performance tends to deliver low delay variability (packet jitter), which in-turn facilities proper operation of mobile hand-offs within access cells (sector-to-sector) or between access cells. This functionality is one of the cardinal/fundamental enablers for the support of converged fixed-mobile services. Many Ethernet network deployments are undertaken without proper consideration of delay and delay variability and as a result present the operators with performance difficulties when high-value, delay-sensitive services/applications are deployed. The Wireless Backhaul Layer: Licensed vs Unlicensed Backhaul Technologies The metro WiMax backhaul layer generally needs to extend the attributes of the metro fiber PoP to which it is typically backhauling traffic to. The general function of the backhaul layer is to extend the metropolitan geographic reach of the metro core network �. since WiMax access is all about getting the wireless signals to within close proximity of the end-sites. As such, the key Ethernet backhaul layer attributes are: High performance traffic policing Bandwidth scalability Low delay and low delay-variability Availability Although these attributes are different, they are effectively co-dependent upon the underlying over-air radio layer. To illustrate this, consider the use of 5.8 GHz unlicensed technology for backhauling WiMax access hubs/cells. In the event that there is interference on critical network links, a number of inter-related things happen in the backhaul network, for example: Interference causes the link to suffer poor RF Signal-to-Noise-Ratio (SNR), driving the link to reduce transmission speed through adaptive reduction in modulation complexity needed to combat the compromised SNR This leads to a reduction in available bandwidth The reduction in available bandwidth leads to escalation in delay due to the higher buffering employed due to the escalated bandwidth-mismatch between the network links and the over-air RF bandwidth Buffering causes escalations in delay and increases delay variability. Additionally, packet losses can result due to buffer over-runs Reductions in network service performance cause service failures and/or outages (also known as 庁navailability�) When the network is focused on the transport of delay-sensitive, high value traffic (i.e. VoIP, VIDoIP, TDMoIP) there is little/no elasticity in the traffic to accommodate the above scenario � so there are usually direct, measurable/noticeable resulting impacts on service quality (i.e. dropped calls, severely-errored seconds, etc). Interestingly when the traffic is highly elastic (i.e. best effort in nature) one might think that the above network conditions might be acceptable. However, higher layers in the applications, when confronted with unsuccessful communications resulting from packet discard (buffer overruns) and inappropriate delays often begin retransmitting �. Potentially further aggravating the network bandwidth problem. 5.8 GHz1 Interference problems are seen by some as non-issues which can be ignored. However, when interference in the unlicensed bands is present it can be very difficult and costly to debug and/or remedy. In an effort to determine the extent of interference as it applies to directional backhaul links that would normally be deployed in elevated scenarios running between metro roof-tops (i.e. over the normal low lying building clutter), field studies2 have been undertaken to record activity in typical large (pop> 5M), mid (pop ~ 1M) and small (pop <> Data sampled3 from hundreds of arbitrary distributed sites within the respective general metropolitan areas indicated that close to 100% were exposed to interference which would be deemed harmful to directional PtP radio link performance4 (see sample in Figure 4 and Figure 5 below). The longer range and high bandwidth demands of PtP backhaul make this system highly susceptible to interference effects as compared to their short range, Multi-Point (MpT) counterparts using the same spectrum. Figure 4� Typical RF Activity Recording � Large Metro Figure 5 - Typical RF Activity Recording � Small Metro In contrast, licensed technology employed in the backhaul layer provides more predictable available RF conditions since [unpredictable] interference conditions are largely non-existent. Using licensed technology, normally at higher frequencies where larger bandwidths are available and where coordinated link congestion is avoided drives the network design to consider rain-induced availability impacts. The network design benefit with this is that the statistical rain databases are available and have been used to successfully predict high frequency radio link availability for many, many years. Today there are literally millions of these links in service globally. The use of licensed backhaul technology provides the foundation for attaining network performance attributes consistent with the basic goals outlined throughout this paper, namely; High Availability and predictability Low, predictable delay and low associated delay-variability Multi-services support, including fixed & mobile applications Notes 1 -- even more so in the 2.4 GHz unlicensed band. 2. Field testing and data analysis conducted by the Canadian Federal Government, Industry Canada, Communications Research Center (CRC). Funded by DragonWave. 3. Data gathered at fixed sites using calibrated, elevated, directional receiver system. 4. Typically levels of 20 � 40 dB of noise floor escalation was measured across the entire 5.8 GHz UNI band, resulting from unknown 5.8 GHz transmission signals received at the test locations.