Exploding the myth of WLAN performance
When most people think of WLAN performance, they think of the 11 Mb/s throughput of 802.11b or the 54 Mb/s throughput of 802.11a or 802.11g. But there’s a lot more to WLAN performance than the theoretical maximum throughput of the radio channel. In fact, the theoretical maximum throughput of a channel has relatively little to do with the way the WLAN performs in a large organization, and it could well be the last thing IT managers should consider when evaluating different WLAN architectures.
Performance on wired Ethernet networks
“Performance” in the networking business relates to the quality of the end user experience. For example, performance of voice calls is always associated with the quality of the call—VoIP calls are considered good when they compare equally with traditional phone calls.
In the wired networking industry, performance is often measured by bandwidth. This follows a decade or more of bandwidth-based arms races between incumbent Ethernet network vendors. Each vendor defined its products as offering better performance because they could carry more bandwidth or deliver higher throughput. As it developed, the nature of wired Ethernet enabled a Moore’s Law-like increase in bandwidth that far outpaced the needs of applications that ran on these networks. It was generally believed that if you could supply enough bandwidth, applications could perform adequately without concern for the “randomness” that is associated with Ethernet, and that if the applications worked well, users would be happy. Hence, more bandwidth meant better performance.
Why WLAN performance is different
After they had defined Ethernet performance as bandwidth for more than 10 years, there was a natural inclination for network equipment vendors to define wireless LAN (WLAN) performance in the same way. But WLANs involve radio transmission and potential interference, and the quality of the user experience can be affected as much or more by issues like access availability or handoff delays as it can by the amount of raw bandwidth available. To understand the difference and see how it affects performance criteria, we need to look at the 802.11 architecture a bit.
802.11’s design was based on an Ethernet hub-like Media Access Control (MAC) model. As developed in the 1990s, the model assumes that the typical WLAN deployment is one, stand-alone access point (AP) for wireless data traffic, either in a small business or in a corporate hot spot such as a conference room. And in fact, such a system provides a fairly high-quality user experience, as long as there’s only one AP and there are only one to three users connected to it.
But WLAN deployments have evolved into multi-AP networks that span floors or buildings. Unfortunately, the initial 802.11 specifications didn’t anticipate the multi-cell deployments or high user densities that are typical in large-scale environments like those found in enterprises.
Pervasive deployments require new assumptions about WLAN performance
The difference between a network designed to serve a hot spot and a network designed for multi-AP deployments has far-reaching implications for measuring true performance beyond speed and throughput. As it turns out, many of the assumptions upon which 802.11 was built prove difficult to sustain in a multi-AP environment.
For example, contention among clients was assumed to be low in a single-AP system, because it was assumed that perhaps five users at most would be connected to the AP at a given time. Since contention was low and VoIP hadn’t yet risen to prominence as an application, there was no need for quality of service control on each transmission—users could connect randomly to the AP, and any delays caused by contention wouldn’t be severe enough to affect the performance of data applications such as web browsing or e-mail. In addition, a single-AP environment meant there was no need to worry about channel overlap or co-channel interference among multiple APs. Also, the single-AP design meant there was no need to worry about handoffs from one AP to another.
For consumer applications, these conditions have remained the same (single AP environment and simple applications) and the rule of thumb (increased bandwidth means better performance) still holds true. But with a need for broader coverage and the rise of time-sensitive VoIP applications driving enterprise WLAN sales, enterprises now need many access points to provide continuous wireless coverage for voice and other application mobility.
When we move from a single-AP or hotspot deployment model to a pervasive, multi-AP deployment model, the assumptions about WLAN architecture and performance must change.
Three cornerstones of WLAN performance
When we assume a multi-AP deployment, a denser user population, and very time-sensitive applications such as VoIP, there is a new definition for performance as it relates to the end-user experience. In order for the end-user experience to be satisfactory, there are three key requirements that did not exist under the standalone “hot-spot” assumptions:
- Density and guaranteed access –
access to and throughput from the network must be maintained regardless of the size of the user population
- QoS –
service quality (and bandwidth) must be guaranteed bi-directionally for high priority traffic (VoIP)
- Coordination -
neighboring APs on the same channel must have clear, coordinated communication so as to not interfere with one another or with clients. The network must also be able to hand off clients between APs without delays.
With the meteoric rise of 802.11 and the millions of compliant products that have shipped to date (thanks to Intel and other low-cost 802.11 vendors) there is one other important design requirement that, while not explicitly a performance issue, still drives product designbackward compatibility with the 802.11 standard. Infrastructure vendors designing products to meet these new requirements must achieve these three cornerstones of performance without any alteration to end-user devices.
Performance limitations with current MAC designs
While the speed of the 802.11 radio is important, these new performance criteria are more important because in many cases, the network won’t work at all unless they are satisfied. Unfortunately, most current access points use the same 802.11 MAC that was designed to meet a different set of assumptions, and thus they cannot meet the new performance requirements. Here’s a closer look at the issues in detail.
Density and guaranteed access
Current 802.11 MAC implementations achieve a peak capacity of about 50 percent efficiency with 3 or more connected users (or contending stations). Additionally, as the number of contending stations increases, aggregate capacity drops precipitously (to less than 1 Mb/s with 10 contending stations). This is due to the overhead of 802.11 transmissions, its use of CSMA/CA (carrier sense multiple access with collision avoidance) with BEB (binary exponential backoff) as the core channel access scheme, and the fact that the transmission medium (air) acts as a very large “Ethernet hub.” As more and more clients try to gain access to the air and transmit packets, more and more collisions ensue. As a result, traffic collides frequently, and stations remain in a “backoff and wait” state more then they actually transmit.
Under the early 802.11 assumptions (low client density and stand-alone “hot-spot” operation) this wasn’t a problem. But when we bring the AP into an enterprise with many offices and users, APs quickly achieve a 10- and even 20-user count within their coverage range. Throughput tails off because of these collisions, and access becomes completely unpredictable because of the random backoff timers.
Due to declining efficiency as user density increases, WLAN users in a densely populated network based on legacy MAC designs will find a starved channel, slow application response and (in the case of VoIP), poor connection quality. Notably, the user experiences poor application response or call quality regardless of the nominal speed of the network – because the density of the clients governs the network’s behavior, the user’s access and call quality would be poor even if there were an 802.11 radio capable of 10 Gb/s speeds.
The new assumption that there will be many APs co-located in the same building presents a very different deployment model for which standard 802.11 MAC was not designed. Coordination is a completely new capability, one that any access point built on consumer grade technology (i.e., the current implementation of the 802.11 MAC) will never be able to support.
In a typical WLAN based on legacy 802.11 MAC technology, two neighboring APs (meaning two APs which can “hear” each other on the same channel – whether or not they are adjacent) can and will interfere with each other’s transmissions. Without coordination at a packet level across APs, transmissions from two different APs and all their users would disrupt each other’s transmissions even if one were using a high priority application. Hence the transmission is not clear for every connection, the quality of the connection suffers, and the user again considers the network to have poor performance. While some manufacturers allow network managers to adjust AP power levels to compensate for interference, it doesn’t completely eliminate the problem.
Additionally, with the presence of multiple APs in a mobile environment, inter-AP handoff becomes a necessity. As users roam the enterprise with their laptops or phones, they will expect that the application “will just work.” Unfortunately, with APs based on current MAC implementations, inter-AP handoff is a complicated “hand-shake” negotiation that is initiated by the client. This has two affects on performance in a multi-AP deployment – unpredictable delays in packet transmission during the negotiation process; and the possibility of every client trying to connect to the same AP as users roam, thereby creating a density imbalance among APs that drops performance to virtually nothing.
Higher WLAN performance requires a new product architecture
What is needed is a complete, 802.11-compliant redesign of WLAN product architectures that implements the 802.11 MAC in a way that facilitates satisfying these new requirements. To find a working model of such product architecture, one need look no further than their pocket or purse - the cellular phone network.
Cellular phone networks were designed from the ground up with the assumption that many phones were going to be used everywhere, and that there would be many base stations (cell towers). Additionally, they were designed specifically for the very strict requirements of the most delay-sensitive application in use today – voice. In other words, the cellular network was designed from the ground up with the same assumptions and performance requirements that we have outlined for WLANs – access to the channel regardless of density, guaranteed signal transmission for voice, and coordination across base stations so quality of the transmission is not affected by co-channel interference or handoffs.
Ultimately, the final evolution of WLAN architecture will incorporate the best of the cellular world (coordination for the ability to manage co-channel interference and handoffs, control of access to the channel to guarantee access regardless of density, and the ability to guarantee prioritization of voice traffic both to and from the end user), and the best of the 802.11 WLAN world (solid data performance and compatibility with millions of existing compliant end points). While centralization of management functions is a first step in addressing enterprise WLAN deployment by simplifying management and security tasks, this architecture doesn’t go far enough because it is still based on consumer grade 802.11 MAC technology and as such has not been designed from the ground up for true WLAN performance in a pervasive, multi-AP environment.
This new architecture would allow enterprises to deploy a true, high performance communication network fully compatible with standard 802.11 technology that satisfies end user’s needs in two ways – data performance is up to par, and voice quality will be as good as or better than that in a cellular phone.
Joel Vincent is Product Marketing Director at Meru Networks.
Visit Meru Networks
Want to use this article? Click here for options!
© 2013 Penton Media Inc.
In this Webinar you will learn how to create a real-time relationship with your customers, how to proactively improve the customer experience, and how to successfully target and cross-sell services to boost incremental revenue.
- Megabytes to Megabucks, Bandwidth to Business Models: How 4G Is Changing Everything
- How to Unplug Your Redundant Telco Apps To Save Money and Improve Efficiency
- When IaaS Isn't Enough: Service Provider Business Models to Drive Growth and Build Margin
- How to Transform Your Aging Telco Voice Network to Drive New Profits and Revenue
- Creative Licensing Approaches for Telcos & Their Network Equipment Vendors
- Smart Home Opportunity: Balancing Customer Data & Privacy
This paper discusses the rise of Diameter and benefits of Diameter Protocol.
- Conducting The Orchestration – Order Management at the Speed of Business
- Toward a Converged Network Edge
- Beyond Spam – Email Security in the Age of Blended Threats
- 6 Important Steps to Evaluating a Web Filtering Solution
- The Expertise to Protect You from Botnet and DDoS Attacks
- Seeing is Believing – Bridging the Order Visibility Gap
Service providers are under tremendous pressure to turn up new services faster then before and, at the same time,
to do it at less expense - and intra-office fiber is one of the biggest challenges in terms of both cost and service
From the Blog
Join the Discussion
Get more out of Connected Planet by visiting our related resources below:
Connected Planet highlights the next generation of service providers, as well as how their customers use services in new ways.Subscribe Now