Upcoming Events

Cloud Connect
Santa Clara
Feb 13-16, 2012

Cloud Connect brings together the entire cloud eco-system to better understand the transformation we're experiencing and promises to be the defining event of the cloud computing industry. Learn about the latest cloud technologies and platforms from thought leaders in Cloud Connect’s comprehensive conference.

Register Now!

More Events »

Subscribe to Newsletter

  • Keep up with all of the latest news and analysis on the fast-moving IT industry with Network Computing newsletters.
Sign Up

FEATURE STORY

The High Wireless Act

by Dave Molta and Josh Linder

A networking world without wires means no punch-down blocks, no cabling systems to document and a new-found mobility for laptop computer users. Is it here today?

It depends. Some technology really does free you, in certain circumstances, from the physical restrictions of wired networking. However, there are tradeoffs like price, distance and performance.

We tested seven wireless LAN products in our Syracuse University labs and on-site in a newly constructed five-story office/classroom building on the Syracuse University campus. In the lab, we measured performance under optimal conditions. In the office/classroom building, we tried to measure real-world performance with an emphasis on distance considerations.

Radio-based products included Solectek Corp.'s AirLAN, Digital's RoamAbout, AT&T's WaveLAN and Windata's FreePort. InfraLAN Wireless Communications' InfraLAN, IBM's Infrared Wireless LAN Adapter and Photonics Corp.'s Collaborative Wireless network system are based on infrared optical transmission technology.

The Truth About Performance

Only InfraLAN could be considered a wired Ethernet replacement if high bandwidth is a major consideration, although its point-to-point architecture is really better suited to interconnecting LAN segments. The rest of the products don't come close, although Windata did turn in some fairly respectable numbers, even under a heavy load.

Most would argue that raw performance is somewhat overrated and that these products pro vide adequate performance for many applications.

Down to the Wire

AT&T, Solectek and Digital offerings are based on radio technology originally developed by NCR Corp., now a part of AT&T. The PC Card adapters these vendors sent us were identical, except for minor differences in drivers and installation programs. Digital and Solectek access points, or hubs, were technical equivalents, except for form-factor and packaging.

Photonics and IBM codevelop and cross-license products. While we were impressed by these systems, we discovered some serious driver problems with the Photonics PC-Card workstation drivers during our multistation testing. The total aggregate throughput for four stations was substantially lower than for a single station. Photonics expects to resolve these problems soon. IBM's offering came with newer drivers for the PC-Card adapter and turned in more respectable multistation performance numbers. Users will find these products appealing for establishing ad-hoc conference-room or classroom LANs.

Digital's RoamAbout narrowly edged out Solectek and Windata offerings as our top choice. Digital's Access Point was the best-designed product, and the support for DES encryption and alternative radio technologies (WaveLAN and Proxim) differentiated it from Solectek. Digital is also the clear leader in providing roaming technology. Windata does well if you are willing to trade portability and roaming for performance.

DEC RoamAbout

Digital provides OEM versions of the WaveLAN PCMCIA adapters and also resells AT&T's ISA adapters. Installing the RoamAbout PCMCIA drivers was similar to AT&T's and harder than it should be. The documentation was clear, but the lack of an index was a notable omission.

Digital adds value primarily through its access point, which Digital designs and manufactures. There are several features that distinguish it. First, it's based on a faster processor and is therefore slightly faster than the AT&T product. It includes a serial console port and accompanying cab le that make configuration much simpler.

It also uses the same PC Card wireless adapter used on client stations, so it's potentially easy to upgrade if the radio technology is enhanced. In fact, by the time you read this, Digital will likely have announced compatibility with Proxim's 2.4-GHz frequency-hopping wireless system.

Finally, and perhaps most important to existing Digital network hub customers, you can easily snap the RoamAbout access point into the DEChub 90 and 900 concentrators to add wireless capabilities to a network with minimal effort.

Solectek Corp. AirLAN

Solectek is one of the most visible players in the wireless LAN market with a wide range of products. Except for its parallel-port Wireless LAN adapter, Solectek's technology is based on OEM products from AT&T and Digital. However, Solectek does a superior job of packaging and support, and its installation utilities and documentation were much better. Wireless networking is Solectek's only business and the company's commitment to this market was evident in our conversations with them.

Solectek's access point is a small device and the installation made it clear that it shared a common heritage with the Digital product--a very good idea. Moreover, it is superior both in design and performance, compared with the AT&T offering. However, Solectek's access-point uses an earlier version of Digital's bridging code that does not support DES encryption. Solectek says it is adding this support. At $2,495, Solectek's access point costs about $700 more than the same box from Digital, but it is packaged with a PC Card radio adapter necessary for operation.

Solectek's parallel-port adapter worked well in our testing, providing performance that was just slightly slower than the PC Card adapter. Some sites may appreciate the flexibility and ease of setup that comes with a parallel adapter.

Although we were disappointed with the compatibility problems we experienced with our IBM ThinkPad 720, we credit Solectek for clearly labeling the boxes with this compatibility information.

Windata Corp. FreePort

Unlike the AT&T, Digital and Solectek troika, which uses spread-spectrum technology in the 902-MHz to 928-MHz range, Windata chose to use the 2.4-GHz and 5.8-GHz frequency bands. While the radios are more expensive to produce, the resulting product offered superb performance, particularly in our multinode tests.

There are tradeoffs for all this power. First, we were fairly disappointed with the distance coverage in our real-world testing. We connected at a maximum distance of just more than 100 feet, and we were unable to communicate with the access points from some locations less than 100 feet away. Windata reported that our building had an unusually high concentration of metal and reinforced concrete obstructions. While we are fairly certain that strategically locating access points throughout the building could have provided full coverage, it would prove both time-consuming and expensive. Windata does provide excellent signal measurement and positioning tools.

Windata's products are also bulky. There is no PC Card form-factor interface. Judging from the heat dissipated from our wireless antenna modules, you shouldn't count on seeing one soon. This product is ideally suited for interconnecting multiple wired nodes into a wireless cluster, and Windata even provides 10BASE-T concentrators for this.

In effect, you're able to construct a network that includes an aggregate of wired and wireless connections consistent with the physical limitations of your facility. The side benefit of this clustered approach is that individual nodes run standard Ethernet, so you can use the same drivers as you do with your wired network.

However, there is potential for interference between wireless products like FreePort that operate in the 2.4-GHz range and "leaky" microwave ovens, which operate in the same spectrum.

Finally, while Windata's product is the most expensive, you get extremely high-quality equipment. It has an unsurpassed fit and finish.

IBM Infrared Wireless LAN Adapter

IBM's Infrared Wireless LAN Adapter has significant potential for low-to-moderate bandwidth applications and is a good value. It uses diffuse infrared technology that bounces signals off ceilings and walls, allowing coverage up to 900 square feet of unobstructed space within a walled area. Although the signals will not pass through walls and its 1-Mbps data rate was the slowest of the products tested, we were still impressed by this ingenious and well-engineered system.

We installed the IBM system after the Photonics Collaborative system, which is virtually identical, so we found client driver installation straightforward. Unlike Photonics, IBM has a single driver for the PC Card adapter, and it worked on all four of our notebook computers without any tweaking. Performance was also significantly better than Photonics', both for single-station and multistation tests.

IBM stands out by offering a software-based access-point at no charge. Every IBM Wireless LAN Adapter ships with simple software bridging code that can turn any PC with an Ethernet card and ISA Infrared adapter into a MAC-layer bridge that runs as a TSR under DOS. Ethernet card support is limited to IBM, SMC, Novell and 3Com, although it is likely you can make other cards work. You can also substitute Token-Ring for Ethernet.

We found installation on our Dell Optiplex 450 awkward. Our initial attempts to install the bridge using a Novell NE-2000+ had to be aborted because the install program couldn't find the right drivers on our NIC driver disk. We eventually got it to install cleanly using a 3Com 3C509. The heart of our problem was that the bridge requires NDIS drivers for your cards, but the documentation is not explicit in that regard. We were also disappointed that the wireless NIC only supported two IRQ options: 10 and 15. While we're certain you could get this to work in most PCs, there's really no excuse for limiting the installer's flexibility this way.

Besides the obvious advantage of low c ost, this TSR bridging approach provides flexibility, since there is no need to dedicate a machine to this function. However, running a bridge as a TSR is not something we would recommend. Not only must the bridge code compete for resources with other applications running on the machine, but a lockup of that machine will likely interrupt network services for the wireless users. If you don't need advanced bridge functionality, this may suffice, however.

Of course, you can dedicate a PC to the bridge function. An IBM engineer quoted us a packet-filtering rate of 10,000 packets per second while running dedicated on a 486 DX2-66 processor.

We were disappointed by the relatively primitive management capabilities and the rather weak documentation. Those reservations aside, we think this is a solid product for users who want to create a single-room ad-hoc wireless network at a very reasonable cost.

AT&T WaveLAN

AT&T's WaveLAN is based on spread-spectrum radio technology originally developed for defense applications. Data is encoded using a direct sequencing technique and transmitted between stations over the nonlicensed 902-MHz to 928-MHz radio spectrum using a variation of Ethernet known as Carrier Sense Multiple Access with Collision Avoidance.

The use of publicly available spectrum is advantageous, since there are no licensing requirements. Still, the fact that the same spectrum is used by some cordless phones and many garage-door openers may be a little unsettling to some. However, given the amount of bandwidth used by those products, it is unlikely that they will cause any significant degradations to network performance. You also cannot assume you're free to use these products outside the U.S., where spectrum restrictions may vary from country to country.

The highlight of the WaveLAN product line has to be the wireless PC Card adapter. It's a type-II PC Card. A short cable attaches the adapter to a radio/antenna module that can be attached to your notebook using a self-adhering plastic bracket. Wh ile no match for our PC Card Ethernet cards, our subjective impressions while running a single wireless node were very positive. We could definitely get some work done with this network. Unfortunately, the WaveLAN PC Card adapter draws considerable power, so be prepared for as much as a 25 to 50 percent reduction in battery life under heavy network I/O conditions.

Although we were able to set up the WaveLAN PC Card adapters and access point without significant difficulty, AT&T could learn a thing or two from Solectek. We did run into problems trying to install the PC Card adapters on our IBM ThinkPad 500 and 720 Notebooks. We subsequently discovered that the PC Card chipsets on some IBM notebooks are incompatible with the WaveLAN adapter.

In addition, the PC Cards use enabler technology rather than PC Card socket services. Since the enabler is a real-mode DOS device driver loaded via the CONFIG.SYS file, the card must be inserted in the PC Card slot when the system is booted to function. Since Windows 3.1 doesn't support hot-swappable networking, this will probably pose only a minor inconvenience. The WaveLAN ISA is supported in the Windows95 M8 beta.

Like most of the other products, the AT&T access point operates as a MAC-layer bridge, allowing WaveLAN nodes to be integrated into an Ethernet network with little difficulty. Alternatively, you could install a WaveLAN ISA card directly in your file server. The AT&T access-point is well constructed and includes a bracket that lets you securely mount it to a wall. AT&T bridge performance was slightly inferior to similar products from Digital and Solectek. Its installation was complicated by the lack of a console port that forced us to configure the unit in-band.

InfraLAN Wireless Communication InfraLAN

The InfraLAN product is unique, offering excellent performance for a limited-distance, line-of-sight, point-to-point environment. If that sounds like too many qualifiers, you feel the same way we do. Nonetheless, this product is a high-performance, low- cost solution that may prove ideal for network environments that don't require a hub-and-spoke topology. InfraLAN provides ATM, Token-Ring and Ethernet-compatible systems.

Unlike others, InfraLAN does not use a hub-based topology, although the vendor plans to introduce such a product soon. Instead, it provides a high-speed connection between two clusters of network nodes, or between a single cluster and an existing wired hub or server.

InfraLAN consists of an optical unit that is normally mounted either on a wall or a tripod, and a transceiver with built-in Attachment Unit Interface (AUI) and 10BASE-T ports. The transceiver can be attached to either a single node or another network hub. For our testing, we attached one InfraLAN transceiver to a Synoptics 2813 workgroup 10BASE-T concentrator and the other to a Synoptics 3030 10BASE-T concentrator to which our server was attached. All of our test nodes were attached to the 2813 concentrator, and the InfraLAN acted essentially as a cable replacement between the hubs. Performance was excellent.

While documentation is rather skimpy, setup was fairly easy. LED indicators help you aim the optical units to establish a connection. This aiming becomes particularly important as you extend the distance toward its maximum of 90 feet, and you also must avoid aiming the optical devices directly at reflective surfaces.

InfraLAN is incapable of penetrating solid obstructions, a point we found easy to prove as we passed a manual between the line of sight of the two optical transceivers. We also experienced some problems setting the units up for direct 10BASE-T connections. The AUI ports worked the first time.

We had trouble getting through to InfraLAN, receiving prerecorded messages during normal business hours. Once we got through, the InfraLAN staff was knowledgeable.

Photonics Corp. Collaborative
Wireless Network System

Photonics' Collaborative Wireless Network System is identical in most respects to IBM's Infrared Wireless LAN Adapter. The PC Card adapt ers were physically indistinguishable, except for labeling, confirmed by the fact that we could run Photonics' PC Card adapter using IBM's drivers. Photonics covers all the bases for network interfaces, including a low-cost ISA adapter, two PC Card adapters (one with a tethered optical unit and the other with an integrated one) and a parallel-port adapter. While we limited our evaluation to PC products, you can even get Macintosh versions that operate at LocalTalk speeds. The most notable difference distinguishing Photonics' product offering from IBM is the parallel-port adapter.

Photonics' access point is a wireless-to-Ethernet bridge made up of a 386-33 PC with 1-MB RAM, a floppy-drive, and Ethernet and Collaborative network interfaces, packaged in a rather handsome case. If you're willing to do a little integration on your own, you can save a few dollars by purchasing a Collaborative ISA adapter and the software separately, and installing it on your own PC.

Setup was easy, but we ran into a minor snag when we tried to follow the installation directions and make a copy of the boot floppy. For some reason, Photonics ships the software on a low-density diskette. No configuration was required--we just inserted the boot floppy and turned on the power and the bridge was functional. SNMP management is not supported.

Installing the PC Card adapter was not nearly so smooth. The documentation wasn't very clear about the proper ODI MLID driver for our system. However, after a little bit of trial and error, we got things up and that's where some fun began. The only way we could get a stable connection to our NetWare 3.12 server was by tipping our notebook on its edge--not a very acceptable situation.

After more experimentation, we discovered that the notebook was extremely sensitive to positioning. After looking through the documentation, it finally hit us that our lab could not possibly have been much worse an environment for this technology. We have large windows, lots of shelves along the walls, boxes stacked f loor-to-ceiling and racks full of computers. We finally decided to move our testing to a conference room measuring approximately 22 feet by 22 feet, and what a difference it made. We could position the computer at any point in the room and still maintain a strong connection to the server. We could even place our hand just above the optical device and it still kept on ticking.

Performance was about what you'd expect from a 1-Mbps system--roughly 10 percent of the speed and capacity of Ethernet when using a single node. Unfortunately, we ran into unexpected degradation as we added nodes. In fact, average aggregate performance for our four-station test, which increased with all of the other products, actually decreased by more than 20 percent. Photonics indicates that the problems we experienced were related to driver problems that are being corrected.

Dave Molta is director of network systems at Syracuse University. His Internet ID is djmolta@syr.edu. You can also e-mail Dave directly . Josh Linder is a network consultant at Syracuse University. He can be reached at jslinder@syr.edu.


Standards: Balancing Simplicity And Flexibility

Although wireless products have begun to make significant market penetration, many users are holding off until standards-based products become available. The Institute of Electrical and Electronics Engineers (IEEE) 802.11 committee has been working since 1990 on a series of physical and MAC-layer standards for spread-spectrum and narrow-band radio as well as infrared technologies. The IEEE is expected to settle on three general physical-layer standards, including a frequency-hopping protocol operating in the 2.4-GHz spectrum, a direct-sequence protocol operating in the 2.4-GHz spectrum and an infrared standard.

Ratification of those standards is mired in the usual vendor politics. Some committee members favor a simple standard and others prefer a more flexible and long lasting--but also more complex--standard. Optimists su ggest that standards are around six months away, but 12 months seems like a more realistic time frame. The original draft standard, which was released for balloting following the November 1994 meeting, was ultimately rejected.

The IEEE 802.11 committee has also been discussing possible standards for transceiver power management, data encryption and data compression.


How Far Can You Go With Wireless LANs?

We tested each of the products in our Syracuse University Lab and, where appropriate, on-site in a newly constructed five-story building. Our distance testing for the optical products took place in our laboratory facility, an adjacent hallway and conference room.

For radio-based products, we installed access-points in a second-floor wiring closet and wandered throughout the building, testing in numerous locations. Overall, we were fairly disappointed in the distance tests. For the WaveLAN-based products (AT&T, Digital and Solectek) we achieved acceptable performance to a distance of 160 feet from the access point.

For the Windata product, the farthest distance we could communicate was just more than 100 feet. After consulting with the vendors, we concluded that the existence of significant amounts of steel reinforced concrete and metal cabinetry had a significant adverse effect on distance.

Performance for each product was measured in both single-station and four-station environments using a combination of Toshiba, Canon and NEC notebook computers. The access points were connected to an isolated Ethernet network in our lab, with performance measured between the notebook computers and a 486-25 server running NetWare 3.12. We used Novell's Perform3 testing utility, which measures performance using a wide range of packet sizes. While useful for controlled comparisons, you should not necessarily assume you will see similar numbers. We performed the Ethernet as well as InfraLAN and Windata tests on the same notebook systems with Novell NE-4100 PC Card adapters.


Roaming Is Not The Sam e As True Mobility

For many potential users, wireless is synonymous with mobility. Unfortunately, while vendors have made progress in incorporating "roaming" features into their products, the current generation of technology is still far from where it should be.

AT&T, Digital, Solectek, IBM and Photonics all include simple roaming capabilities that allow notebook users to move between access points without disturbing any connections to servers. In essence, the access points hand off the signals in much the same way a cellular phone system does. Unfortunately, what is not supported in standard product offerings is roaming across network boundaries. Thus, users who try to roam between buildings with LANs that have different IPX network addresses will need to reboot their systems to reestablish a connection to the network.

The same general rule applies to TCP/IP subnetworks, although Digital does offer a product called RoamAbout Mobile IP that solves the problem using a roaming-enabled TCP/IP client stack, a Mobile Device Server installed on each subnet, and the Dynamic Host Configuration Protocol (DHCP), which dynamically allocates addresses. To alleviate any problems that might result from security capabilities built around IP addresses, RoamAbout Mobile IP lets users maintain their original IP address even when they cross subnet boundaries with the Mobile Device Server acting as a translator.

Unfortunately, while we commend Digital for its efforts, we don't think it'll have too many takers with the current offering. Not only does it require the installation and administration of dedicated PCs on each subnet to act as device servers, it also ties you to a DOS-based TCP/IP stack at a time when most of the world is migrating toward Windows VxD protocol stacks. While rumors suggest Novell is close, many sites won't be able to consider this kind of solution until they deliver roaming capabilities for NetWare and IPX.

Research and Reports

Hypervisor Derby
August 2011

Network Computing: August 2011

TechWeb Careers