The Need for QoS in Wireless Networks

An additional network requirement must be supported if the user experience in wireless broadband is to be similar to the user experience in wired broadband (for example, T1 access). The previous paragraphs detailed how wired IP networks can be engineered to limit latency and other factors that detract from QoS. The IEEE has been grappling with the issue of QoS in wireless networks and has recently approved 802.11e, which is backward compatible with other variants of 802.11. That means that improvements in QoS contained in 802.11e can be applied to 802.11 or 802.11a. This section outlines the mechanisms used to ensure the QoS contained in both 802.11 and 802.11e.

Challenges to Wireless QoS

Many previous attempts at WLAN QoS (and non-QoS channel access schemes) show that the strategies that work well in a wired environment do not translate to WLAN. Several factors break possible assumptions: The packet error rate can be in the range of 10 to 20 percent, bit rates vary according to channel conditions, and the "rubber pipe problem" arises where bandwidth managers do not know how much bandwidth they have to manage because a neighboring, unrelated bandwidth manager can take some of it at any time. In addition, if a wireless network is to bypass or substitute for the PSTN, it must be able to prioritize voice and video packets over data packets.^[24]

Latency in Wireless Networks

As witnessed earlier in this chapter, the chief threat to an IP network is latency (or delay) of the delivery of packets via the network. Latency is defined as the time it takes for the network to respond to a user command. If latency is high, causing noticeable delays in downloading web pages, then the experience feels nothing at all like broadband, no matter how high the data rates are. Low latency (less than 50 ms) is a requirement for the successful mass-market adoption of wireless services and devices.

The latency experienced by wireless users has a number of contributing sources, including the air link processing, propagation, network processing and transport, the far-end server (if applicable), the application, and the user device. The sum of these latencies must be minimized to ensure a positive end-user experience. Because of the many contributing sources in wired networks, there is little room for latency contributed by the wireless system. Table 5-6 describes the types of delay encountered on an 802.11 network.

Table 5-6: Types of delay encountered on an 802.11 network

Delay	Definition
Air link processing	The time necessary to convert user data to air link packets (code, modulate, and frame user data) and transmit it
Propagation	The time necessary for a signal to travel the distance between the base station and the subscriber device, and vice versa
Network transmission	The time necessary to send the packet across the backhaul and backbone networks, including routing and protocol processing delays and transmission time
Far-end processing	The time required for processing by the far-end servers and other devices
Source: Flarion

The processing delay leads to another very unique disadvantage of systems that are not built on an all-IP basis. Many networks cannot transmit native IP packets and require IP assistance through either a protocol change (transcoding and encapsulation) or the addition of equipment in the network to simulate IP performance. Those measures introduce complexity and packet delays, further impacting the latency of a given system and driving up cost.

Throughput and latency are two essentials for network performance. When taken together, these elements define the speed of a network. Whereas throughput is the quantity of data that can pass from source to destination in a specific time, round-trip latency is the time it takes for a single data transaction to occur (that is, the time between requesting data and receiving it). Latency can also be thought of as the time it takes for data to be sent on one end to be retrieved on the other end (from one user to the other).

Latency is crucial to the broadband experience because the Internet is based on the Transmission Control Protocol (TCP). TCP requires the recipient of a packet to acknowledge its receipt. If the sender does not receive a receipt in a certain amount of time (ms), then TCP assumes that the connection is congested and slows down the rate at which it sends packets. TCP is very effective in dealing with congestion on the wired networks.

A system's ability to efficiently handle a large user population depends significantly on its ability to service many small TCP/IP messages per unit time and multiplex many active data users within a given cell. Hence, high latency translates directly into lower system capacity for serving data users, which equates to higher cost. The ideal mobile data network supports both high peak data rates (greater than 3 Mbps) and low packet latency (less than 50 ms). A unique approach is needed to do this over the wireless medium.^[25]

^[24]A presentation from Intel.

^[25]"Low Latency—The Forgotten Piece of the Mobile Broadband Puzzle," a white paper from Flarion, www.flarion.com.