Improving WLAN Performance with Fragmentation
September 23, 2002
Learn how the optional fragmentation function of 802.11 operates, and understand when and how to use it.
In order to make the best use of wireless LAN bandwidth, you need to be aware of various 802.11 configuration parameters. As an optional user feature, the 802.11 standard includes the ability for radio-based network interface cards (NICs) and access points to fragment packets for improving performance in the presence of RF interference and marginal coverage areas. Let's take a closer look at how fragmentation works and cover some implementation tips.
Fragmentation in Action
A source (NIC or access point) uses fragmentation to divide 802.11 frames into smaller pieces (fragments) that are sent separately to the destination. Each fragment consists of a MAC Layer header, frame check sequence (FCS), and a fragment number indicating its ordered position within the frame. Because the source station transmits each fragment independently, the receiving station replies with a separate acknowledgement for each fragment.
An 802.11 station only applies fragmentation to frames having a unicast receiver address. This includes any frame (e.g., data frame) directed toward a specific station. In order to minimize overhead on the network, 802.11 doesn't fragment broadcast (e.g., beacons) and multicast frames.
The Sequence Control field of each fragment header includes a Fragment Number subfield, indicating the fragment number of the frame. The number is zero for the first fragment, then increments by one for each successive fragment of a particular frame. The single-bit More Fragment field in the fragment header indicates whether or not a frame is the last of a series of fragments. The More Fragment field is set to "1" if the source station will be sending additional fragments of the same frame. It's set to zero if no more fragments will follow.
The destination station reassembles the fragments back into the original frame using fragment numbers found in the header of each frame. After ensuring the frame is complete, the station hands the frame up to higher layers for processing. Even though fragmentation involves more overhead, its use can result in better performance if you tune it properly.
Fragmentation Implementation Tips
The use of fragmentation can increase the reliability of frame transmissions. Because of sending smaller frames, collisions are much less likely to occur. The fragment size value can typically be set between 256 and 2,048 bytes. This value is user controllable. In fact, you activate fragmentation by setting a particular frame size threshold (in bytes). If the frame that the access point is transmitting is larger than the threshold, it will trigger the fragmentation function. If the packet size is equal to or less than the threshold, the access point will not use fragmentation. Of course, setting the threshold to the largest value (2,048 bytes) effectively disables fragmentation.Similar to RTS/CTS, a good method to find out if you should activate fragmentation is to monitor the wireless LAN for collisions. If you find a relatively large number of collisions, then try using fragmentation. This can improve throughput if the fragmentation threshold is set just right.
If very few collisions (less than 5 percent) are occurring, then don't bother. The additional headers applied to each fragment would dramatically increase the overhead on the network, reducing throughput. That, you want to avoid.
If significant numbers of collisions are occurring, try setting the fragmentation threshold to around 1,000 bytes first, then tweak it until you find the best results. After invoking fragmentation, follow-up with some testing to determine if the number of collisions is less and the resulting throughput is better. You should try a different setting or discontinue using it altogether if the throughput drops (even if you have fewer collisions).
The use of 802.11 simulation tools (e.g., Opnet) can aid in determining optimum fragmentation threshold sizes, but you'll need to develop a simulation model that mimics your network. It will be difficult to accurately portray the RF interference on the actual network. As a result, fine-tune the actual WLAN in conjunction with real testing results.
As a final note, keep in mind that the use of RTS/CTS could be a better way to reduce collisions if hidden nodes are present. It's best to jointly consider the use of RTS/CTS and fragmentation before settling on which one to use. As with any 802.11 tuning mechanisms, the goal is to improve performance. If what you do improves throughput, then you're doing the right thing.
Join Jim for discussions as he answers questions in the 802.11 Planet Forums.
Ready to do your site's survey? Join us at the 802.11 Planet Conference & Expo, Dec. 3-5 in Santa Clara, CA. Author Jim Geier will be there in a workshop covering "RF Site Survey Basics: Equipment and Techniques for Maximizing WLAN Performance ."