|
| |
Designing Metro Mesh Architectures for Maximum Applications Performance
It should be clear at this point that user expectations for metro-scale WiFi meshes are very high, and likely to increase further as these networks become increasingly mission-critical for all three categories of users noted above. The need to support both current and emerging applications requiring greater speed and performance motivates increased attention on the architecture of the products used to construct such networks.
We discuss a number of key attributes that assist in defining production-quality metro-scale WiFi mesh networks. We believe that application demands will only increase over time, underscoring the need for certain key architectural features that we believe will become core requirements.
A significant attribute to metro-scale wireless mesh networks is capacity, or more precisely - the ability handle potentially large amounts of data, including time-bounded voice and video traffic for potentially many simultaneous users and applications at any given location, but also the ability to grow and scale economically and cost-effectively.
We have found that there are three key attributes of a WiFi mesh architecture that address these goals, as follows:
Ability to provision sufficient capacity at any given point in the mesh
The demands on networks of any form only increase over time, as the number of users, their transmit duty cycles, the size of the data objects they seek to transfer, and the degree of time-boundedness of an ever-greater percentage of those objects all grow. Meshes are in one respect no different from any other network – bottlenecks absolutely will have a detrimental effect on throughput and capacity. Thus it behooves the designers of metro-scale WiFi meshes to carefully consider the alternatives in addressing these concerns.
Unlike most other networks, the subscriber capacity of a mesh actually increases with each additional infrastructure node added. What differentiates one mesh architecture from another is the efficiency with which it transports subscriber traffic from a given point over a number of hops to either the destination within the mesh or a point of interconnect to backhaul. Some mesh architectures are extremely efficient in carrying subscriber traffic over many hops and can also require fewer backhaul connections than other implementations as a result. Less-efficient implementations can also require more mesh nodes than might otherwise be required, increasing costs and lowering total cost of ownership and thus return on investment. The key architectural difference that defines efficient mesh architectures is the ability to provision a larger number of radios per node. More radios yield more capacity.
Note that the use of multiple radios includes support for both subscriber access and intra-mesh interconnect. A larger number of radios dedicated to each increases overall capacity, as we noted above, but it is also important that these two services be balanced so as to avoid congestion and blocking. Note also that more radio per node can minimize the requirement for additional (and usually expensive) off-mesh backhaul capacity bridging to external networks. The links can be provisioned less frequently in the mesh, minimizing costs again.
Mesh algorithms – These are the “secret sauce” in all modern mesh implementations, as mesh-equipment vendors continue to work diligently to devise protocols with the right combination of throughput, resilience, and intelligence in adapting to multiple classes of service with highly-varying and usually unpredictable instantaneous data loads. Such protocols must also handle mobility, an enormous challenge in and of itself, and always have a key goal of minimizing both intra- and inter-node latency. Strix DMA™ is an excellent example of the industry's highest performance mesh algorithm enabling increased capacity, scalability, range and mobility. It is important to note that mesh architectures and their related protocols fall into two key categories - routing (or Layer 3 protocols, referring to layer 3 in the International Standards Organization reference-model description) and switching (Layer 2 protocols). Vendors of Layer-2 implementations make a convincing case for greater efficiency via the lower inter-nodal (and perhaps intra-nodal) overhead in this approach. In general, inter-node latency needs to be below 5 milliseconds (ms.), while aggregate inter-node latency over multiple hops should not exceed 50 ms. for time-bounded traffic
Backhaul capacity – Finally, we define backhaul here as the connection(s) between the mesh and external networks. Backhaul connections are typically provided using either point-to-point radio, point-to-multipoint radio or fiber. Dual- and single-radio mesh architectures may require substantially more individual backhaul links due to the lack of intra-mesh capacity required to reach a given backhaul connection. All this, of course, further motivates the provisioning of multiple radios per node as the preferred architectural strategy.
There are a number of additional architectural attributes that we believe are also important:
Centralized management – A requirement for medium- and especially large-scale deployments of wireless networks is the ability to centralize provisioning and troubleshooting and provide graphical monitoring, event logging, and control for potentially hundreds of infrastructure nodes and thousands of radios.
Field upgradeability – While mesh nodes tend to be inexpensive, they become much more cost-effective when they can be provisioned with multiple radios and especially when these additional radios can be added in the field. This eliminates the requirement for a wholesale “forklift upgrade”, as it is known within the industry, and helps to minimize capital expense (CapEx).
Security – While security must be largely addressed as a network, and is not just a wireless, concern, it is important that wireless meshes implement two key features. The first of these is support for all WiFi Layer-2 security techniques (WPA and/or WPA2), and the second is full support for user-directed encryption and authentication, including VPNs of various forms and 802.1x/EAP or other user-specified and -provisioned authentication.
Multiple classes of service – As we noted above, it’s important to be able to provision multiple classes of service (and classes of users) within a single mesh network. Each of these can have different traffic prioritization levels, security, and pricing profiles as may be required. This capability can have a profound impact on both cost minimization and potential revenue opportunities. Typically, such facilities are provided as independent VLANs, each associated with a separate SSID. It is important to note that for truly independent classes of service with separate encryption protocols and authentication mechanisms each SSID must have a unique MAC address.
High Speed mobility – Finally, as we noted above, we believe that high-mobility applications involving moving vehicles will become much more important over time. The key element here is the ability to process an inter-node handoff for rapidly-moving clients in less than 50 ms. In addition, as we noted above, the ability to maintain session connectivity at speeds exceeding 200KM/hour is essential for commuter rail applications.
For more information on this topic click here
|
|
|
|