The Future Of Wi-Fi

Wireless networking used to be seen as complex and awkward, requiring arcane knowledge of radio engineering as well as network protocols. Wi-Fi made it seem much simpler--perhaps even too simple, as anyone who has tried to hunt down a rogue access point (AP) can attest.

It's about to get complicated again. The well-known IEEE 802.11a and 802.11b technologies have been joined by an alphabet soup of variants, today stretching as far as 802.11v. Most of these newer standards haven't been finalized yet, but they all represent innovations required before Wi-Fi can truly be considered mature and enterprise-class. In the meantime, buyers have to negotiate an array of competing proprietary approaches.

The coming innovations go beyond the well-trodden grounds of security and QoS. While both are important, they're problems that vendors have already addressed. Many Wi-Fi networks aren't secure, but that's because of poor configuration or obsolete equipment, not because of weaknesses inherent in the products or standards themselves. Thanks to 802.11i and Wi-Fi Protected Access (WPA), wireless networks can be as secure as their wired counterparts.

The same is true for QoS. Although the 802.11e standard for traffic prioritization and bandwidth provisioning hasn't been formally ratified yet, most new data products support it, and the Wi-Fi Alliance is certifying interoperability for the parts of it that deal with prioritization under its Wi-Fi Multimedia (WMM) program. In the first few months of 2005, we'll see Wi-Fi phones that support both the QoS and security standards.

In fact, the greatest room for improvement is in network management, with new technologies emerging for both the radio link and the network infrastructure. Smarter transceivers could eliminate the need for radio engineering expertise within a year. The signaling protocols required for roaming between APs and from Wi-Fi to 3G will take longer, slowly emerging during 2006.There are new developments on the software side, too. The integration of physical location tracking with virtual presence databases could enable people to be reached at any time via the most appropriate communications method.

The move to higher data rates will be more gradual, with 802.11n products not due to ship until 2007. However, many vendors are offering technology this year that will improve the performance of existing clients. The result will be a growing number of truly wireless LANs and WANs that use Ethernet only for linking switches to APs, not to clients.

The Self-Configuring Network

Cheaper and more intelligent APs are eliminating the need for manual site surveys. For two years, Wi-Fi vendors have argued about where best to embed 802.11 intelligence. Should it go inside each individual AP, or be relegated to a central appliance? The answer is now becoming clear: To cope with the ever-changing radio environment, both are necessary.

The next generation of APs will be smart enough to take independent action when necessary, yet still cooperative enough to work together and follow instructions from air traffic control. But it doesn't stop there. For optimum performance, tomorrow's Wi-Fi network management will also need to enlist client devices.

SLIGHT SURVEY

Many network architects are already benefiting from distributed, cooperative Wi-Fi intelligence. Its initial effect has been to reduce the importance of the site survey, the time-consuming and expensive process of determining where to place APs.

Vendors such as Airespace and Trapeze Networks sell systems that can perform their own site surveys in real time, automatically adjusting the transmission power of existing APs as coverage requirements change. By continuously monitoring the airwaves, they can also help detect intruders and track down rogue APs, avoiding the need for a technician to walk around with a spectrum analyzer.As APs get smaller and cheaper and Power over Ethernet (PoE) becomes widespread, the need for a site survey could disappear entirely. Instead of designing a network to use as few APs as possible, network architects can simply put APs everywhere and let them organize themselves. Aruba Wireless Networks is already promoting this concept as a "wireless grid," a network that combines wired and wireless connectivity by including a Wi-Fi AP inside every Ethernet wall jack.

FORMATION FLYING

Better coordination of APs will also reduce interference, which in turn increases data throughput and improves reliability. Meru Networks and Extricom both sell Wi-Fi gear that can use the same radio channel for every AP in the network, eliminating frequency planning and theoretically tripling the airwaves' capacity. Cisco Systems has concentrated on the client side, persuading almost every NIC manufacturer to include Cisco Compatible Extensions (CCX) in their hardware. CCX allows Cisco Aironet APs to gather data about the radio environment from clients, then fine-tune transmissions to suit each one.

All these technologies are currently proprietary, only working with APs from a single vendor, but two emerging standards should change that. On the infrastructure side, Airespace and Nokia are leading an IETF effort to develop a way for different vendors' APs to work together, known as Control and Provisioning of Wireless Access Points (CAPWAP). The group hopes to have an RFC ready by May 2005. However, interoperable products that support the protocol could still be a year away, and Cisco hasn't committed to support it yet.

Interoperability is even more important on the radio side because most networks include a wider variety of clients than APs. The IEEE's newest wireless project, 802.11v, is essentially the same as CAPWAP, but the institute has already standardized a potentially more important specification, known as Transmit Power Control (TPC). TPC includes some of the same radio management functions as CCX, but adds the ability for clients to reduce their own transmission levels--something that can also increase battery life.TPC is part of the 802.11h standard, intended to satisfy European regulators worried about wireless networks interfering with satellite communications. As such, not all Wi-Fi equipment intended for the U.S. market supports it yet, but it will become increasingly prevalent during 2005.

signal strength: real-time radio management

Many organizations have already abandoned traditional site surveys, but the vendors and standards bodies still have work to do before clients can manage themselves. The need to support legacy 802.11b hardware and deal with non-802.11 interference will slow progress further.

Presence Gets Real

Location-aware networks can link your physical position to virtual availability information.If convergence is ever to become more than just a buzzword, communications will have to be routed based on presence--that is, data that specifies how you can be reached, and whether you want to be.

Most presence data still consists of the same kind of information found on IM buddy lists, such as whether a person is on the phone or offline. Wi-Fi will make it much more powerful by adding information about the person's physical position.

RADAR TRAP

Today, the simplest way to locate a user is to find out which AP that user is connected to--information that every network already has. This isn't very accurate in a traditional network of ceiling-mounted APs that each cover hundreds of employees, but it improves as the AP density increases. If you're putting an AP in every cubicle, it's likely to be good enough.

The next step is to use triangulation, comparing the signal strength received from several APs. Most vendors offer this capability as part of their automated management software, using it to direct wireless coverage toward areas with the highest user density. However, triangulation doesn't work as well with Wi-Fi as with cell phones: The 802.11 protocol isn't designed to let clients connect to several APs simultaneously, and interference in unlicensed frequencies means that signal strength can vary erratically.Building on triangulation, RF fingerprinting compares signals against a pre-existing radio model of the network. This can be accurate to within a few feet and is particularly good at determining position relative to physical barriers such as walls--something particularly useful for security purposes.

So far, only two vendors offer RF fingerprinting. Airespace builds it into its own APs, while wireless security specialist Newbury Networks includes it as part of a separate management overlay that works alongside other vendors' APs. The technology's biggest weakness is that it's only as good as the initial radio model, which needs to be calibrated through an old-fashioned manual site survey.

The whole point of presence data is to route communications via the most appropriate technology, so location can also be tracked using one of these other technologies. For example, Nokia's converged Communicator device combines GPS with GSM voice and Wi-Fi data. Thanks to E-911 mandates, many cell phones also include GPS, and so will many of the Wi-Fi phones expected in 2005. GPS provides accurate 3D coordinates, but works better outside than in buildings.

CONVERGING CONNECTIONS

Getting all this technology to work together will be difficult. Enterprise presence databases usually reside within IP PBXs, which can't yet store or process location data from Wi-Fi networks. The only vendor that makes both Wi-Fi gear and IP PBXs is Cisco, though IP PBX leader Siemens is buying Wi-Fi start-up Chantry Networks.Others rely on partnerships. Avaya is already integrating its IP PBX with APs from Proxim and Extreme Networks and says location-aware presence is an important part of this. Multivendor interoperability may arrive by 2007, thanks to SIMPLE, the IETF's emerging XML schema for storing and transferring presence data.

signal strength: LOCATION TRACKING

Location tracking has been held back by competing proprietary technologies, but is already possible. Even without further innovation, denser networks will make the radio side more accurate, while regulation will force the development of better software.

Roaming Meets RoutingTrue mobility requires interoperability between different standards, not just between rival vendors.

Signal Strength: Seamless Roaming

Roaming is the hardest problem to solve in Wi-Fi, requiring the cooperation of many competing vendors and standards bodies. For the foreseeable future, it will require proprietary overlay technologies.

Fans of 802.11 telephony like to talk about a single device that can roam between public and private networks, smoothly transferring both voice calls and data sessions from a 3G cellular network to a Wi-Fi network. This will eventually be possible, but not this year, and perhaps not next year either.

The problems are both commercial and technical. The carriers that subsidize cell phones currently have little incentive to support roaming onto Wi-Fi. On the Wi-Fi side, the industry still hasn't settled on a way for clients to move between different vendors' APs, let alone from Wi-Fi to cellular.SWITCHING CELLS

The trend away from standalone APs and toward centralized switches has helped simplify roaming. From the client's perspective, no roaming is necessary while moving between APs attached to the same switch. However, unless the network is very small or the switch very large, users also need to roam between switches.

With the notable exception of Cisco, most enterprise-class Wi-Fi switch vendors do allow connections to be transferred between switches--provided they're all made by that particular vendor. Non-proprietary roaming will require 802.11r, a still-vague standard unlikely to arrive before 2006.

It isn't just switches and APs that need upgrading to improve roaming. In a dense network, client devices may be able to see several different APs, so they need a way to decide which one they should connect to. Another proposed standard, 802.11k, aims to help with this. It specifies ways for clients to gather information about the signal strength from various APs, then sends the information back to a switch, which determines the most appropriate connection.

The 802.11k group published its first draft specification in March 2004, but since then its members have reached a dead end on exactly what type of data clients should be reporting back to switches. Meanwhile, proprietary solutions to the same problem have emerged.The most widely supported of these on the client side is Cisco's CCX. The most advanced is a technology developed by Propagate Networks, which is built into Atheros' chips. These are used by most enterprise APs, but clients are more likely to use Intel.

BACK TO FORWARDING

The cell phone industry hasn't been ignoring Wi-Fi. In September 2004, a group of companies led by Nokia announced an initiative known as Unlicensed Mobile Access (UMA), which aims to make GSM and GPRS services available over Wi-Fi and Bluetooth networks.

UMA doesn't entail roaming so far. Its most important feature is call forwarding. The calls may travel over a Wi-Fi network, but they're still routed through the cellular operator, so will still show up on the phone bill.

Antennas Everywhere802.11n's predecessors are already here, but don't get trapped by proprietary systems.

Signal Strength: Smart Antennas and MIMO

802.11n will happen, and the IEEE's target of late 2006 seems realistic. The only hurdle to overcome is the choice between the rival proposals, which are so similar that a compromise should be possible. But watch out for proprietary "Pre-N" equipment.

Officially, nothing has been announced about the high-speed 802.11n standard, except that it's at least two years away and will mean Wi-Fi networks that reach speeds of 100Mbits/sec or higher. Unofficially, almost everyone in the industry agrees how it will work, and some vendors are already shipping "Pre-N" equipment.

Network architects should be skeptical of the official silence and even more skeptical of vendors that claim compatibility with a non-existent specification. While the core technologies set for inclusion in 802.11n are no secret, the IEEE's goals are so ambitious that a lot more innovation is required.DIFFERENT PATHS

The actual throughput of most Wi-Fi links is a small fraction of what vendors claim. That's because performance degrades with distance, so clients usually need to be within 100 feet of an AP to reach the full data rate of 54Mbits/sec. The airwaves are shared media, so running a full duplex link immediately halves that rate, and adding more than one client device reduces it further. The 802.11 protocols' high overhead halves that number again.

802.11n aims to be different. It's designed to provide 100Mbits/sec of real TCP/IP throughput over a range of at least 100 feet. This means that the raw bit rate--the number we'll soon see quoted in vendor hype--has to be at least 500Mbits/sec, and possibly much higher.

There are several ways to achieve these high speeds, but most vendors agree on two: Use more radio spectrum, or use more antennas. Using more spectrum is simpler, but it's inefficient. If every network segment uses twice as much spectrum, only half as many network segments can share the same airwaves. For this reason, 802.11n will likely limit each segment to two 802.11 channels and require the uncrowded 5GHz band.

Adding antennas is more efficient, thanks to Multiple In, Multiple Out (MIMO), a technique that sends different signals via different spatial routes and can even boost performance with existing hardware. The disadvantage is that extra antennas cost more and take up space, so 802.11n will likely include a slower option for smaller, cheaper devices.MIMO MOBILITY

This past summer, the IEEE's member vendors submitted 61 different proposals for 802.11n. However, by October they had coalesced into just two alliances: World-Wide Spectrum Efficiency (WWiSE) and TGn Sync. Both have very similar proposals. The main difference is that WWiSE wants to preserve the existing 802.11 MAC, whereas TGn Sync wants a newer, more efficient one. A new MAC would make equipment more complex and expensive, but boost speeds even further.

TGn Sync is the clear leader. It's larger and has heavyweights Intel and Cisco on board. But WWiSE might still win out. It's led by Airgo Networks, the company that invented MIMO and is currently shipping "Pre-N" chips that use three antennas to double the data rates of 802.11a and 802.11g. TGn Sync member Atheros says it will ship its own version of MIMO in the summer of 2005, but hasn't yet said how many antennas it will use, or what the data rate will be. However, the two alliances' proposals are so similar that they should be able to resolve their differences by 2006.

The End of Ethernet?As Wi-Fi gets faster and more reliable, wire will retreat to the network core.

Signal Strength: Wi-Fi Mesh Networks

The Ethernet NIC might eventually go the way of the floppy disk drive, but bandwidth constraints mean that Wi-Fi alone is only suitable for homes and very small offices. Large networks will still depend on copper to link APs together.

Wi-Fi watchers spent last year arguing about thin vs. fat APs. We're set to spend this year arguing about integrated vs. overlay networks. Is it better to buy Wi-Fi gear from the same supplier as Ethernet switches, or from a vendor that specializes in wireless?

The argument is less important than it seems. No vendor has been designing Ethernet gear to support Wi-Fi for long, so your installed base of Cisco, Extreme, and Foundry switches won't necessarily work any better with switches from those vendors than their competitors. Conversely, wireless-only start-ups see a large proportion of their sales through Ethernet vendors that don't have a wireless solution of their own. The integrated Wi-Fi solutions from 3Com, Alcatel, and HP are really provided by Trapeze, Airespace, and Aruba, respectively.But in the long term, most Wi-Fi vendors don't see 802.11 as an overlay or something to be integrated with Ethernet. They want to replace it entirely.

COPPER OUT

There are two separate visions of an all-wireless network. The least radical simply says users won't bother with wires once Wi-Fi is comparable to Ethernet in speed and reliability. Espoused by vendors such as Symbol Technologies and Airespace, it entails an extremely dense network of APs. Copper will still be needed to link them together, and additional wires may even be needed. In addition to an AP on every desk, you'll need them in the elevator and in the bathroom.

Vendors such as Strix Systems and Vivato propose something more revolutionary: APs that are entirely free of wires, using Wi-Fi for their uplink to the core network, as well as to cover the last few feet.

This is much harder to realize because of bandwidth scarcity. There simply isn't enough spectrum in the unlicensed bands to aggregate several 100Mbit/sec links together. Nevertheless, it will be a useful alternative in buildings that aren't already cabled. There's even a nascent standard, 802.11s, that could eventually turn every client into an AP.An all-wireless future will require more than just Wi-Fi. Devices will need to support other standards such as 3G and perhaps WiMAX, automatically choosing the most appropriate communications method.

The short-range 802.15 technologies will also play a role. Bluetooth is already integrated into most cell phones in Europe, and some U.S. carriers are trialing a location-based presence service based on ZigBee, a low-power radio with a range of only a few feet. Developed by Bridgeport Networks, the service adds ZigBee chips to both landline and cell phones. Whenever the chips are within range of each other, calls to the cell phone are automatically diverted to the landline.

DEAD AS THE DESKTOP

Wireless skeptics point out that wired networks are advancing, too. Many PC vendors now include Gigabit Ethernet NICs as standard, and even 802.11n will be unable to compete with these. Wi-Fi is always a step behind.

The counter to this is that few users will ever need more than 100Mbits/sec. The PC industry has already found that there's little demand for increased speed alone. Rather than trying to make desktop PCs ever faster, vendors such as Intel are increasingly focused on shrinking desktop power into a lightweight and energy-efficient laptop.The networking industry is going the same way. High-end workstations and legacy desktops will still require wires. But for employees whose laptop is their primary or only computer, abandoning Ethernet is a realistic prospect.

Respond to Andy Dornan's commentary on wireless networking at wires.networkmagazine.com.

Overlay Networks

Pros: Can choose best-of-breed wireless vendor * Non-wireless vendors can offer integrated solutions using overlay equipmentCons: Two networks to manage * Feature sets of wired and wireless products may overlap

Integrated Networks

Pros: Management is often simplified * Can purchase complete network from one vendor

Cons: Vendor lock-in is a risk * Third-party solutions may still be necessary if the single vendor doesn't offer all required functionality

For more on QoS and wireless VoIP, see October 2004's "Voice Over Wi-Fi: Too Green For the Enterprise".

Different Wi-Fi switching architectures are covered in November 2003's "Building the Intelligent Wireless LAN".

MIMO is explained in-depth in August 2004's "Boosting Wi-Fi With MIMO and 802.11n."

See January 2004's "EAP: Extending Authentication to the Wireless LAN" for a full description of the latest Wi-Fi security standards.