802.11a tips, tricks and traps

Products with the 802.11a designation are just hitting the market, offering high bandwidth and more channels than 802.11b products, available for about two years. But before you upgrade (or before you go wireless for the first time), there are some tricks and traps to be aware of.

Knowing the differences

If you've had experience with the 802.11b standard in its current form, you know it has three channels available for use. This number is important because it dictates how 802.11b access points are placed to cover wireless geography into areas called cells. 802.11a has eight channels, and therefore while radio broadcast patterns are roughly the same, programming cells are decidedly different between the two standards. Things get even stranger if you're planning to use the dual-channel or hybrid 802.11a/b access points.

If you associate access points as cells, each cell has a broadcast pattern available for client hardware. The client hardware will associate with access points at a specific data rate that is a function of the quality and signal strength between the two devices.

The 802.11a product group is similar to 802.11b in that clients associate with an access point at a data rate where packet error rate is low. The difference is that data rates for 802.11a fall back from a peak of 54M bit/sec through a range of slower data rates -- typically 48M, 36M, 24M down to a minimum of 6M bit/sec. By contrast, 802.11b products we've tested typically fall back to half the data rate of 11M bit/sec at the first step, yielding a typical throughput of about 250K byte/sec. As multiple user demands of a single access point climbs, true throughput can start to crawl with as few as three active users per 802.11b access point.

We've found that a high concentration of users near one access point in 802.11b causes clogging, which is not found in 802.11a. Users close to an access point can tend to dominate the access point to the detriment of those farther away, but both standards can demonstrate this effect if there are lots of high-data rate applications such as streaming media (think CNN downloads) from the same access point. The best cure we've tested so far for this is the bandwidth throttling that's available from the Bluesocket wireless gateway that we tested a few months ago -- and it works with 802.11a and 802.11b.

Behind the scenes, 802.11a has other advantages that let access points provide steadier service. As an example, two techniques in 802.11a help slow fallback and provide stronger throughput. The first is sending redundant data in a process called forward error correction, which isn't present in 802.11b. This gives added immunity to electromagnetic interference and noise that might otherwise corrupt data and cause fallback or retransmission. Packets in the received signal have error correction content with no effective overhead or wasted space. Periodicity per transaction is reduced, increasing overall 802.11a availability, found uniformly across all the 802.11a access points we tested.

The second reason is that 802.11a's signal transmission, orthogonal frequency division multiplexing, has a higher immunity from the effects of multipath signals. These occur when reflections of a signal arrive a fraction of a second before or after the desired signal, causing confusion at the radio receiver in the access point or client card.

Nonetheless, 802.11a data rates can seem to change whimsically. We discovered during tests of a D-Link DWL5000AP and the

D-Link DWL-A520 adapter (the first PCI-bus 802.11a card that we could find), that the data rates would change from 36M up to 54M bit/sec, then drop back again periodically for no apparent reason. The D-Link PCI-bus card, a top performer otherwise in our tests, has an antenna mounted on the rear of the card's backplane. The 'aha' moment came when we discovered that a person walking down the hallway behind the access point temporarily increased the access point's signal -- boosting it to a higher data rate. We couldn't convince the individual to stand in the hallway during major downloads. We also saw that moving file cabinets and large metal objects sometimes had a bearing on reflectivity -- and could even unblock a signal or send it from a nominal speed to a much higher one.

The CCI conundrum

Being able to roam throughout an 802.11a geography is important. Hand-off from one access point's radio toanother ensures uninterrupted connectivity. Some applications, such as Microsoft's Pocket Outlook on an iPAQ running Windows Pocket PC, react rudely when a message download is interrupted during a gap in coverage.

Co-channel Interference (CCI) occurs when users can associate (communicate with) two access points on the same channel. CCI ties up association time, which can be a detriment to other users on either access point. This can happen at perimeters of cells, where the data rates are slowest.

Using power settings to limit the geography covered by an 802.11a or 802.11b access point can reduce the duplicate channel coverage phenomenon somewhat, with a possible sacrifice of overall throughput. A throughput reduction occurs because access point data rates fall as a function of signal, and signal is reduced as a function of distance (and antenna dispersion patterns/ strength) from an access point.

The 802.11a products can reduce this effect because there are more channels available (eight channels, although Atheros now says it can theoretically go up to 13, although it is untested and potentially nonstandard), and therefore a greater distance is allowed between two access points that use the same channel (see graphic, below). Using a combination of power tweaking, antenna positioning and channel selection, a larger cell geography can be deployed without running into CCI problems.

High-density deployments, including those found in public access spaces (coffee shops, libraries and airport lounges) might have some immunity to the problem, mainly because people won't know that there's a problem. Without analyzers, the service-level agreement for access is the grumbling/finger-drumming threshold. Users often won't complain about speed publicly, but will remember privately the responsiveness or perceived throughput that they had.

So far, the client software we've used has not shown a CCI problem. A good addition to client-side drive software would be an indication that a dual/multiple-association or CCI problem is occurring, thus letting a user perhaps move to a better location. The effect is only dramatized by impatience from perceived slowness, similar to a slow modem connection.

Some access points use reverse TNC connectors that let optional antennae be used in 802.11b products, but there's a trend away from using detachable antennae in 802.11a products because of potential conflict in the frequency channels allocated to 802.11a. This potentially thwarts misuse, but also robs those deploying access points of their ability to choose optimal antennae.

We also found that mapping a geography for 802.11a deployment involves a combination of science and magic. While software often lets a map be drawn, it's difficult to know the signal bouncing/reflectivity and blocking values for equipment, wall studs and panels, elevator shafts and other objects that can cut or wreak havoc on the radio signals in 802.11 wireless LANs. For example, we used an 802.11b antenna from Agere Systems, which is said to have a broader dispersion pattern. We tested it first in an open warehouse area, where we found a stronger signal strength (110-feet line-of-sight at full speed, we fell back to a slower speed at 353 feet). We have been unable to find vendors willing to put external antenna connectors on 802.11a products, as Federal Communications Commission and international regulations have higher containment in the 5-GHz spectrum that 802.11a products operate in.

Designing LANs

Most enterprise-oriented wireless LAN products (both 802.11a and 802.11b) come with site survey tools. Typically, the tools will let you define a perimeter geography. Then the tool helps a wireless LAN designer place access points to optimize coverage, although CCI never seems to be mentioned with these tools. CCI is a potential issue with 802.11a, but not like it is in 802.11b; the increased number of available channels in 802.11a helps reduce CCI dramatically, but it's still possible in some installations, especially multifloor/office buildings using older construction, to have interference.

The only useful methods we found to control signal in captive-antenna access points was to experiment repositioning them on horizontal and vertical axes, and by using access point power options to control the diameter of broadcast cell pattern. This is because most all-captive antennae are omni-directional. Using these methods to control dispersion can have a handsome payoff in controlling CCI.

Agere, Intel and others have software that can help with CCI and access point placement. Agere's Orinoco Client Manager Site Manager application allows either 802.11a or 802.11b site roaming/wandering missions (using their cards in a supported notebook) that can show the access points and their relative signal strength/quality, so you can optimize overlapping access point channels. Both vendors' site surveying products used 802.11b samplings instead of actual 802.11a examples; nonetheless both are reasonable at finding and testing access points in cell-mapping missions.

Also, several Linux tools are emerging that replicate the functionality of vendor-specific signal quality monitoring tools, such as KOrinoco (for the KDE UI). We've used this tool frequently when conducting multivendor 802.11a interoperability and monitoring tests.

The fate of the duals

Last month at NetWorld+Interop 2002, vendors (including Intel, Cisco and Agere) announced dual-technology access points. Dual-technology devices will initially amount to two radios in a common base access point device. Some, such as Orinoco's AP-2000 and Cisco's Aironet 1200, were announced as evolutionary products that can accommodate 802.11a and 802.11b technologies. Both vendors, however, said these products also might accommodate slots for upcoming 802.11g products (Click for more on 802.11g).

The primary benefit of these hybrid devices is wireless LAN compatibility and cost savings for companies that have 802.11a and 802.11b networks. However, the cost of the compatibility might cause other theoretical problems. While radios for each standard have essentially no interference with each other, deploying dual/hybrid access points also means channelizing the access points effectively.

Remember, the limitation on the number of available channels for 802.11b products means that proper cell overlaps and coverage might increase the incidence of CCI for 802.11b products, which cannibalizes available bandwidth. The complexity of correctly channelizing and optimizing will become tougher.

Dual-mode cards will likely become a popular method to ensure the ability to connect to whatever wireless LAN resources are available in an area. But this also creates an interesting conundrum: Will the settings default to the fastest connection, or will defaults become enforced by deployment and therefore organizational policy? It is unadvisable for PCs to have two IP addresses concurrently from a management and security perspective (not to mention Dynamic Host Configuration Protocol and roaming issues), although multihoming is common in statically located devices such as servers-that-route and Web servers.

Security issues: Demand the 'X' factor

Like 802.11b products, 802.11a products usually ship using default settings that vendors say will "improve the initial customer experience." Which means they're at the lowest setting or not on at all.

Settings are most critical at access points, because access points are linked to internal network resources, and Internet resources if available through internal network resources.

The security problems surrounding 802.11b are well known. Rogue access points are often poorly configured for security and might permit traffic that can be hard for intrusion-detection software to pinpoint. In turn, rogue access points defeat security perimeters in ways that drive network security analysts into early retirement.

The 802.11a product groups natively use the same Wired Equivalent Privacy (WEP) security that 802.11b products have, and are therefore vulnerable to cracking tools. Some vendors, such as Orinoco and Proxim, have included natively configurable (albeit non-standard) high-encryption capabilities into their access points to prevent simple WEP cracking. Efforts to include 802.1X security into products to allow Remote Authentication Dial-In User Service server-supplied session keys also will pay off, although implementation is spotty and interoperability largely untested (see our iLabs report).

Final thoughts

Part of the quality test of the products that we've seen surround enterprise management and integration potential. Proxim's Harmony product line represents the strongest enterprise-class product for a number of reasons, not the least of which is that wireless LANs created under Harmony guidelines have the best trade-offs we've seen for ease of deployment vs. installation/ integration and security features.

Products from vendors such as Bluesocket and ReefEdge provide strong gateway and bridging and wireless LAN security boundary capabilities for any of the 802.11 product groupings. And while we've tested and liked Bluesocket WG-1000 features, we also wish that the strong security sentiments it represents were the first item seen in every 802.11a product box. Connecting alone isn't enough; securely connecting should be mandatory by default from every box. A mad dash to take advantage of unbaked technology is what's given e-mail, Web and other security problems that hamper deployment and give computing a bad rep as a result.

Wi-Fi and Wi-Fi5

If you ask the Wireless Ethernet Compatibility Alliance (WECA), they'll tell you there's an easy way to tell the difference between 802.11b and 802.11a. The 802.11b products are certified to a standard called Wi-Fi-Certified, and shortly (once there are two different 5-GHz chipsets) the 802.11a products will be certified to a standard known as Wi-Fi5-Certified. Both certifications speak strongly to interoperability, but there is no interoperability between 802.11a and 802.11b without the use of dual-technology products.

There are currently more than 320 products with Wi-Fi certification. When the Wi-Fi5 program begins, there is a backlog of 90 products hoping for certification.

Henderson is principal researcher for ExtremeLabs, of Indianapolis. You can contact him at thenderson@extremelabs.com.

Global Test Alliance

Henderson is also a member of the Network World Global Test Alliance, a cooperative of the premier reviewers in the network industry, each bringing to bear years of practical experience on every review. For more Test Alliance information, including what it takes to become a member, go to www.nwfusion.com/alliance.

Audio primer: Wireless LANs
Wireless LANs can make it easy to support a roving workforce, but pitfalls abound. In this 8-minute primer we take a look how wireless LANs are setup, the challenges involved and the varied specifications underlying the technology. Network World Fusion.

Tech Insider: Wireless LANs
Wireless LANs can provide network flexibility and end user mobility. But first you must decide which flavor of 802.11 technology to select. We spell out the options as well as take a look at security concerns and what pieces of the puzzle you'll need to complete your network. Network World, 05/20/02.

Early users laud 802.11a LANs
The first class of 802.11a wireless LAN users say the technology lives up to its advanced billing as a much speedier alternative to its predecessors, although it is not without shortcomings. Network World, 05/27/02.

Down and dirty with Wireless LAN security
iLabs team puts 802.1x standard to the interoperability test and brings deployment issues to light. Network World, 05/06/02.

What is 802.1x?
Understanding what the IEEE 802.1x standard is and why you should care means understanding three separate concepts: PPP, EAP and 802.1x itself. Network World Fusion, 05/06/02.

802.1X provides user authentication
This mobility has created a need to identify who is attempting to gain access to a given port. The 802.1X standard provides such a solution. Network World, 03/25/02.