Both 802.11a and 802.11g offer data rates up to 5 times faster than 802.11b. But 802.11g sales leap-frogged the slightly older 802.11a, due to backwards compatibility with legacy 802.11b products. Unlike A, which occupies the 5 GHz UNII band, B and G share the 2.4 GHz ISM band, letting G APs simultaneously communicate with B and G stations.
However, as many WLAN owners have noticed, there is a price to be paid for backwards compatibility: the presence of B stations can degrade G performance. To get the most from your 802.11g AP, it can be helpful to understand the "protection" mechanism that enables B/G coexistence at the expense of performance.
Why is protection needed?
G achieves higher throughput using a more efficient modulation called Orthogonal Frequency Division Multiplexing (OFDM). For backwards compatibility, G products also support the older Direct Sequence Spread Spectrum (DSSS) modulation used by B products. According to the 802.11g standard, Extended Rate Physical (ERP) OFDM stations can use data rates up to 54 Mbps. Vendor-proprietary techniques like channel bonding can achieve higher rates, but that's irrelevant for standard B/G compatibility.
Speaking both OFDM and DSSS is a good start, but it is not enough to enable peaceful coexistence. A radio channel is a shared medium; transmitters must cooperate to avoid collisions. To "take turns speaking," all 802.11 stations implement a standard Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) protocol.
When an 802.11 station has data to send, it listens to the channel. If another 802.11 node is transmitting, the station waits until the channel is free. Compare this to meeting attendees who politely try to avoid speaking when someone else has the floor. In a large meeting, you might raise your hand and wait to be called upon. In CSMA/CA, this is done by exchanging short Request-to-Send /Clear-to-Send (RTS/CTS) packets between a station and AP when data exceeds a threshold. RTS/CTS adds considerable overhead, but can sometimes increase throughput by eliminating lengthy retransmissions that would result if "hidden nodes" sent data simultaneously, colliding with each other.
Now imagine that you're in a meeting in a noisy conference center in Madrid. Many attendees are fluent in Spanish, but you only speak English. You're having a hard time differentiating between background conversations and questions posed by other attendees. To overcome this, the moderator tells everyone to ask their questions by speaking slowly into a microphone in English, a language common to all attendees.
The corresponding 802.11g mechanism is called Protection. Since B stations can't understand CSMA/CA over OFDM, G stations must speak a little CSMA/CA over DSSS – just enough so that B stations can tell when G stations are using the channel. When they have data to send, G stations inform everyone sending either an ordinary RTS packet or a CTS-to-Self packet over DSSS. A CTS-to-Self packet is an ordinary CTS packet that carries the transmitter's own address. With RTS/CTS, the sender asks for permission, and the AP tells everyone how long the sender can use the channel. With CTS-to-Self, the sender just announces to everyone that it plans to transmit.
Why protection impacts performance
In any WLAN, RTS/CTS should be used sparingly to overcome "hidden node" problems (i.e., where stations hear a common AP but cannot hear each other). In a mixed B/G WLAN without "hidden nodes," using RTS/CTS all the time would be overkill. CTS-to-Self adds less latency, but still incurs noticeable overhead. Therefore, G products may accomplish Protection using a mix of RTS/CTS and CTS-to-Self, only when Protection is actually required.
In an infrastructure WLAN, it is up to the AP to decide when Protection is needed. G APs announce whether Protection is required by setting a flag in an "ERP Information Element" included in Beacon and Probe Response packets. In the absence of B stations, the AP sets Use_Protection = 0 so that G stations can operate with greater efficiency. If the AP sets Use_Protection = 1, all G stations must immediately begin using Protection mechanisms to politely share the channel with B stations.
As a result, mixed B/G WLANs take an immediate performance hit when the first B station associates. Furthermore, the greater the number of B stations, the larger their impact on WLAN performance. That impact varies, depending upon RF environment, layout, distance, and product, but compare these Planet3 Wireless CWAP Study Guide throughput examples:
-- 22.1 Mbps: 10 G and 0 B stations -- 12.0 Mbps: 10 G and 1 B stations -- 9.4 Mbps: 10 G and 5 B stations -- 8.2 Mbps: 10 G and 10 B stations -- 5.9 Mbps: 0 G and 10 B stations
Although somewhat dated, further examples can be found in this O'Reilly article and these Tom's Hardware test results. Keep in mind that throughput is typically no better than half the max data rate, reflecting total application traffic carried by a shared channel. Close G stations still connect at 54 Mbps, close B stations still connect at 11 Mbps, and data rates for both decline with distance. In other words, G stations are NOT limited to B data rates – the entire WLAN's throughput drops due to Protection overhead (i.e., sending extra RTS/CTS packets over DSSS, with longer interframe spacing).
Minimizing the impact
Vendors can choose how to mix RTS/CTS and CTS-to-Self to implement Protection, influenced by configurable parameters. But, ultimately, 802.11g-compliant stations must follow the AP's mandate to use Protection. 802.11g-compliant APs must set Use_Protection = 1 if any "non-ERP" stations are associated. Protection may also be required if the AP detects an overlapping WLAN (i.e., another Infrastructure AP or an Ad Hoc station on the same channel) that is using only non-ERP data rates, or sending Beacon / Probe Responses with Use_Protection = 1.
Given the large number of laptops with embedded 802.11b adapters, and the proliferation of 802.11b WLANs in homes, hotspots, and small businesses, many WLANs will be forced to deal with B/G coexistence for years to come. So what can you do to minimize B's impact on your G-based WLANs?
Stop buying B-only products. Maybe you can't prevent others from using B near your WLAN, but you can make an effort to migrate your own WLAN. Recall that G performance degrades in proportion to the number of B stations. Do your part to minimize that avoiding B "closeout" sales and upgrading legacy devices.
Configure (some) APs to use G-only. Many APs can be instructed to operate in G-only mode, inhibiting or disabling their ability to support B stations. As you migrate to G, gradually convert newer portions of your WLAN into G-only bastions. Use Protection only where you need it for your own B stations, or to coexist peacefully with nearby B neighbors.
Assign different channels to B/G and G-only APs. Collisions can be reduced by segregating B and G traffic. A G-only AP and a B/G AP on the same channel will cause co-channel interference (noise and retransmissions) which degrades the performance of both WLANs. But a G-only AP tuned to Channel 11 won't be impacted by a B/G AP tuned to Channel 1, beyond brief periods when B stations send Probe Requests on Channel 11. Given limited channels in the 2.4 GHz ISM band, this may only be practical in very small WLANs. Which brings us to…
Consider buying A+G devices. Although backwards (in)compatibility hurt A sales, the 5 GHz UNII band offers less interference and many more channels. New A+G adapters have the best of both worlds: a platform that's compatible with legacy home/hotspot (B/G) APs and new enterprise (A+G) APs. You shouldn't use A just to reduce the impact of Protection, but if you're buying new equipment anyway, it makes sense to consider the benefits of upgrading to A.
- Tune B/G performance parameters. If you encounter excessive retransmissions, you may have a "hidden node" problem that could be improved by decreasing RTS Threshold. Or set your AP to G-only (for example, disable "CTS Protection" on a Linksys WAP54G, or set "CTS Mode" to none on a D-Link XTreme G AP) and measure whether you end up with excessive retransmissions. Determine impact by watching AP counters, monitoring the channel with a WLAN analyzer, and measuring application throughput on several stations.
In the long run, 802.11b products will fade away, replaced by newer 802.11g, 802.11a, and eventually 802.11n products. During the transition, examining how 802.11b may be impacting your WLAN may help you get more Mbps out of your 802.11g products.
|Read about Lisa|
About the author: Lisa Phifer is vice president of Core Competence Inc., a consulting firm specializing in network security and management technology. Phifer has been involved in the design, implementation, and evaluation of data communications, internetworking, security, and network management products for nearly 20 years. She teaches about wireless LANs and virtual private networking at industry conferences and has written extensively about network infrastructure and security technologies for numerous publications. She is also a site expert to SearchMobileComputing.com and SearchNetworking.com.