Fragmentation and Reassembly

Higher-level packets and some large management frames may need to be broken into smaller pieces to fit through the wireless channel. Fragmentation may also help improve reliability in the presence of interference. Wireless LAN stations may attempt to fragment transmissions so that interference affects only small fragments, not large frames. By immediately reducing the amount of data that can be corrupted by interference, fragmentation may result in a higher effective throughput. Interference may come from a variety of sources. Some, but by no means all, microwave ovens cause interference with 2.4 GHz networks.[*] Electromagnetic radiation is generated by the magnetron tube during its ramp-up and ramp-down, so microwaves emit interference half the time. Many newer cordless phones also cause interference.[] Outdoor networks are subject to a much wider variety of interference.

[images/ent/U2020.GIF border=0>] If you need to use a cordless phone in the same area as a wireless LAN, I suggest purchasing a 900 MHz cordless phone on eBay.

Wireless LAN stations may attempt to fragment transmissions so that interference affects only small fragments, not large frames. By immediately reducing the amount

Spectralink Voice Priority

One of the challenges in supporting voice on wireless networks is that voice is far more sensitive to poor network service than data applications. If a 1,500 byte fragment of a graphics file is a tenth of a second late, the typical user will not even notice. If a delay of a tenth of a second is introduced into a phone conversation, though, it will be too much.

Providing high-quality service over an IP network is hard enough. Doing so over a wireless LAN is doubly challenging. One of the major problems that network engineers face in designing wireless LAN voice networks is that all data is treated equally. If there is a short voice frame and a long data frame, there is no inherent preference for one or the other.

Spectralink, a manufacturer of handheld 802.11 phones, has devised a special set of protocol extensions, called Spectralink Voice Priority (SVP), to assist in making the network more useful for voice transport. SVP consists of components implemented in both access points and in handsets to prioritze voice over data and coordinate several voice calls on a single AP. SVP assists with both the downlink from the AP and the uplink from handsets.

To support SVP, an access point must transmit voice frames with zero backoff. Rather than selecting a backoff slot number as required by the 802.11 standard, access points with SVP enabled will always choose zero. In the presence of contention for the wireless medium, the voice frames with zero backoff will have de facto priority boost because data frames are likely to have a positive backoff slot. Strictly speaking, stations implementing zero backoff are no longer compliant with 802.11 because it mandates selection of a backoff slot in accordance with defined rules. (To preserve stability under load, however, retransmitted voice frames are subject to the backoff rules.)

By selecting zero backoff, access points implementing SVP ensure that voice frames have preferential access to the air. Access points that implement SVP must also keep track of voice frames and provide preferential queuing treatment as well. SVP requires that voice frames be pushed to the head of the queue for transmission. APs implement transmit queues in many different ways; the important point is the functional result, which is that voice frames move to the head of the line. Some APs may move voice frames up to the head of a single transmit queue, while other APs may maintain multiple transmit queues and serve the high-priority voice queue first.

of data that can be corrupted by interference, fragmentation may result in a higher effective throughput. Fragmentation takes place when the length of a higher-level packet exceeds the fragmentation threshold configured by the network administrator. Fragments all have the same frame sequence number but have ascending fragment numbers to aid in reassembly. Frame control information also indicates whether more fragments are coming. All of the fragments that comprise a frame are normally sent in a fragmentation burst, which is shown in Figure 3-8. This figure also incorporates an RTS/CTS exchange, because it is common for the fragmentation and RTS/CTS thresholds to be set to the same value. The figure also shows how the NAV and SIFS are used in combination to control access to the medium.

Figure 3-8. Fragmentation burst

Fragments and their acknowledgments are separated by the SIFS, so a station retains control of the channel during a fragmentation burst. The NAV is also used to ensure that other stations do not use the channel during the fragmentation burst. As with any RTS/CTS exchange, the RTS and CTS both set the NAV from the expected time to the end of the first fragments in the air. Subsequent fragments then form a chain. Each fragment sets the NAV to hold the medium until the end of the acknowledgment for the next frame. Fragment 0 sets the NAV to hold the medium until ACK 1, fragment 1 sets the NAV to hold the medium until ACK 2, and so on. After the last fragment and its acknowledgment have been sent, the NAV is set to 0, indicating that the medium will be released after the fragmentation burst completes.