FEATURE

The Bridges of Wireless County

by Dave Molta

Feature Workshop: Crystal Radios and Rabbit Ears

Although fiber is the ideal medium for interconnecting buildings on an extended LAN, if there's no pathway between locations or your warehouse is a mile and a half from your manufacturing facility, you'll need leased serial lines. Typically this includes 1.5-Mbps T1 lines, along with the associated DSU/CSUs and bridges or routers.

In most cases, you're probably looking at one-time costs ranging from $7,000 to $10,000, and recurring costs ranging from $400 to $600 per month. If you could find a way to get better performance for lower one-time costs and no recurring charges, would you be interested?

That's essentially the promise of wireless bridge offerings built around spread-spectrum radio technology. We put six leading products through strenuous testing at our Syracuse University lab and, in general, we liked what we saw. These bridges offer good performance for most applications, they're relatively easy to set up, and, based on our experience with a system that's been installed at our site and survived a couple of Syracuse winters, they're reliable as well.

Persoft's Intersect Remote Bridge and C-SPEC's OverLAN Wireless Bridge are our top choices. While they are based on similar wireless network technologies, they each offer advantages you'll want to consider and weigh against your specific site requirements.

Solectek's AIRLAN/Bridge Plus is similar in most respects to C-SPE C's OverLAN. Nevertheless, it lacked some features and didn't perform quite as well. Aironet's Arlan 640 is an extremely sophisticated offering that packs a lot of punch into a very small form-factor, but the Arlan 640's performance needs improvement.

If high performance is your primary need, and you are able to live with limited distance and a basic feature set, look no farther than Windata's AirPort II, which tripled the performance of any other product. If, on the other hand, distance is the primary requirement, it's tough to do better than Cylink's unique full-duplex offering, AirLink Bridge, although you do trade some performance for the added distance, particularly in environments where network traffic flows primarily in one direction.

Persoft Intersect Remote Bridge

Persoft was one of the first vendors to enter the low-cost wireless bridge market, and while its product isn't glamorous, it definitely gets the job done. Setup is easy, performance is excellent and the feature-set is comprehensive. Its management capabilities are second only to Aironet, and the documentation provides a clear explanation of standard point-to-point and multipoint topologies, as well as clear directions for site planning and installation.

On the downside, you'll pay a little more for the Intersect Remote Bridge than for products with comparable performance and feature sets. Persoft also offers a similar product for Token-Ring sites.

Like the Solectek and C-SPEC products, Persoft's bridge uses AT&T's WaveLAN radio and runs bridge code under DOS. You configure the bridge via a VT-series terminal connection to the bridge's serial port using a null-modem cable. We used the MS Windows Terminal program, but Persoft also supplies a capable DOS-based program. Persoft includes a modem within the bridge for out-of-band configuration and management. No other vendor offers this, but we doubt very many sites would be willing to dedicate a phone line to this purpose.

Like the Windata bridge, Persoft's product boot s from a floppy diskette. While this approach makes you depend on a mechanical component for system software loading, it does provide some flexibility, since the boot diskette can be configured on any DOS workstation by running the CONFIG.EXE program. The configuration program is well laid out, but someday soon we hope one of the vendors will offer a graphical configuration application.

When the bridge PC boots, it presents you with options to configure the software or run the RangeFinder software--a utility that aids in antenna aiming. If no option is selected within 30 seconds, the bridge software loads. Total boot time is approximately one minute, another minor irritant attributable to the floppy-based boot system.

Security and packet filtering capabilities are similar to those offered by other vendors, with DES encryption offered as an option. Persoft supports SNMP MIB II and RFC 1286 Bridge MIB, along with a limited proprietary MIB extension consisting of nine fields related to radio quality and signal levels. While not quite as comprehensive as other offerings, most sites will find it adequate.

C-SPEC OverLAN Wireless Bridge

C-SPEC's OverLAN wireless bridge caused us many headaches. The initial units were configured with a software release that had a major bug and the hardware was improperly configured. C-SPEC resolved the problem and shipped us two new units that performed flawlessly, accompanied by an explanation of what went wrong with the initial shipment and the steps C-SPEC was taking to avoid such an occurrence in the future. On overall technical merit, we preferred the C-SPEC offering to any other, but our bad experience with the initial units casts enough doubt for us to keep it on equal footing with Persoft's.

The OverLAN Wireless Bridge shares many characteristics with the Persoft and Solectek products, foremost of which is its use of the AT&T WaveLAN wireless adapter. C-SPEC one-ups the competition by including a handy array of front-panel LEDs that display signal qual ity, radio and Ethernet data transmission, and bridge forwarding rate. In addition, C-SPEC includes the DES encryption chip--which is optional on the other products--as a standard feature, and it is the only vendor that provides IP routing in addition to bridging. Users bridging Ethernet LANs with very high traffic levels will appreciate the storm threshold capabilities that limit forwarding of broadcast or multicast packets, which can sometimes saturate a low-bandwidth link. The bridge software is virtually identical to Solectek's--a result of both vendors' partnership with a common third-party bridge software developer. But C-SPEC includes additional capabilities like IP routing that Solectek doesn't offer. At least Solectek has some company in that C-SPEC also requires a packet driver for its configuration software.

The C-SPEC configuration and monitoring utility is an easy-to-use DOS application that runs on top of a packet-driver interface. In a day when ODI and NDIS interfaces are the norm, requiring network managers to install a packet driver to run the configuration and management software strikes us as silly. To its credit, C-SPEC provided a software "shim" that let us run the application over our ODI driver. Unlike Persoft and Aironet, C-SPEC does not provide out-of-band management capabilities, though plans are in the works to add this.

While C-SPEC's documentation is thorough, it looks like it was produced by a bad copy machine. It also lacks organization and some of the explanations of features are confusing.

Initial performance statistics had us, and the vendor, scratching our heads. We eventually determined that the units we received were defective and once we retested with new units, performance was near the top.

Reservations aside, this product provides the best combination of ease-of-configuration and installation, bridge/router functionality, management and performance of any of the products we tested. It also does so at a very reasonable cost with support policies that are very ap pealing.

Solectek AIRLAN/Bridge Plus

There's not much to complain about when it comes to Solectek's AIRLAN Wireless Bridge, but there's also not a whole lot to get excited about compared to the competition, either. Built around a stylish PC chassis running bridge code under DOS, we found the AIRLAN Bridge easy to install, packed with features and a capable performer. It has its shortcomings, including an awkward power switch and lack of full configuration options from the console, and it falls short compared with C-SPEC both in features and performance. Nevertheless, its overall value is still solid.

Setting up the AIRLAN was easy. We connected our Ethernet segments, snapped on the antenna leads and powered it up. Attaching a standard PC keyboard and monitor lets you aim the antennas using the utility software. While this system is effective, we preferred the radio signal LED indicators included on the C-SPEC and Cylink products. The monitor screen also reports information on packet throughput and signal quality. However, to alter the default settings, you'll need to run Solectek's DOS-based ABCONFIG software, which also turns a PC into an SNMP management station. While we have no complaints with this application, it does lack out-of-band management capabilities.

Performance was above average overall, but poor for those using AT&T's WaveLAN radio. Security and packet filtering capabilities were fairly extensive, but you'll pay an additional $200 per bridge for the optional DES encryption chip.

The bottom line: Solectek's offering is very similar to C-SPEC's, but lacks some of the latter's features and offers lower performance. However, Solectek has a much broader line of wireless network offerings, including an innovative variation of the AIRLAN Wireless Bridge called AIRLAN/Bridge Ultra. It not only extends wireless distance to 25 miles, but extends the maximum length of cabling between the bridge and antenna--an important consideration for some sites. Solectek accomplishes this by us ing a hybrid 900-MHz/2.4-GHz communication system that combines the benefits of both systems into an offering that's only about 30 percent more expensive than its standard product.

Aironet Arlan 640 Wireless Bridge

In many respects, Aironet's Arlan 640 is the best designed and most sophisticated unit we tested. It is attractive, compact, loaded with advanced features and well documented. Its long-range Telex Yagi antenna was light and compact, making it easy to mount in a concealed location. Its support for multiple channels and unique multipoint configuration capabilities make it extremely flexible. It can even act as an access point for laptops equipped with Aironet wireless PC Card adapters while simultaneously providing bridging capabilities. By the way, it was also the least expensive unit we tested. Unfortunately, its performance didn't measure up.

The documentation was the best of any product we tested, providing clear and detailed instructions for installation, configuration and management. The units can be configured via either a terminal or a Telnet connection. It was the only unit that offered Telnet configuration. The configuration program uses a hierarchical text-oriented interface that we found easy to use, despite an incredibly wide range of options.

The bridge software loads from Flash ROM and there are several options for updating the software, including Xmodem transfers to the serial port, FTP and TFTP. We encountered some initial difficulties applying a software update, but Aironet's technical support diagnosed the problem quickly as an incompatibility with the Windows 3.1 Terminal program. Aironet has since corrected the problem.

The unit is equipped with three multicolor, top-panel LEDs that show wireless connectivity, packet transmission and errors. These LEDs also serve as indicators for link tests, which can aid in antenna aiming. Additional LEDs adjacent to the Ethernet ports indicate which port is active and show packet transmissions and collisions. While we salu te Aironet for its efforts, the LEDs could be consolidated and better labeled to make them more useful. The Arlan SNMP implementation is the most complete and well documented of any of the units we tested, including proprietary extensions that provide full configuration and monitoring capabilities.

The only area where Aironet fell short was in performance. Since the unit is rated at 2 Mbps, we couldn't figure out why aggregate throughput on our file I/O benchmark never surpassed 500 Kbps. Aironet said the Arlan bridge was optimized for flexibility and reliability rather than raw speed. Unlike the WaveLAN-based units we tested, each Arlan radio transmission requires a positive acknowledgment, introducing additional overhead. Still more overhead is introduced by software options that let the bridge be deployed in complex multipoint configurations and as a combination bridge/access point.

Aironet offered two suggestions for sites that need higher throughput. First, up to three Arlan units can run in parallel on unique RF channels, resulting in a raw aggregate bandwidth of 6 Mbps. Second, future versions of Arlan will have options that optimize throughput for bridging applications, including reduced RF header sizes and relaxed RF acknowledgment requirements.

Windata AirPort II Wireless System

Windata's AirPort II Wireless Interbuilding System is as bulky as a 747 and far more expensive than any other unit we tested, but it is also the performance winner by a landslide. Throughput on some tests approached 60 percent of full Ethernet bandwidth! The major trade-offs are installation complexity and distance limitations. The rather bulky AirPort antenna is similar in appearance to a digital satellite receiver, measures 26 inches in diameter and requires a one-inch diameter mounting pipe. The cabling is also heavy-duty. This is definitely not an installation for beginners.

The AirPort bridge had the most restrictive distance limitation of any of the units we tested, with a rated distance of 0.9 mile s. In a multipoint configuration, multiple antennas can be attached to the hub station, thereby extending the total theoretical wireless network diameter to 1.8 miles. We were only able to test it at a distance of approximately one-third of a mile, but at that distance, we experienced no serious problems.

Unlike the other products, the AirPort bridge consists of unique hub and remote units, both of which provide 10BASE-T and AUI, but not 10BASE-2, interfaces. During our testing, we discovered that performance was faster from the hub to the remote unit than it is in the opposite direction. If you are installing a simple point-to-point link, you should carefully consider your network traffic characteristics and plan the location of hub and remote units accordingly.

Software configuration was relatively easy and was accomplished by running a DOS-based application from the supplied floppy disk, which, once configured, acts as a boot floppy for the hub unit. We don't care for the boot floppy, especially for a unit in this price category.

You must properly configure bridge hubs with their companion's name(s) and serial number(s). To enable SNMP support, you must enter appropriate IP parameters along with get and set community names. Although we didn't evaluate it, Windata offers an optional Windows-based SNMP management application called SeePort at a cost of $995. It lets you remotely configure and modify an AirPort network and review network performance statistics.

Bridge functionality is rather limited, with no support for filtering. In addition, Windata recommends disabling the internal bridge code in favor of an external bridge if the number of nodes on your Ethernet network exceeds 256 devices. In fact, by default, bridging is disabled. We think customers spending $20,000 on a wireless bridge deserve full bridge capabilities. Nonetheless, if fast flight time is at the top of your list, AirPort II is the best product.

Cylink AirLink Bridge

Cylink's AirLink Bridge stands out from the othe r products, and for the most part, the distinctions are positive. The most notable difference is Cylink's support for connections at distances up to 20 miles. While we were not able to test the product at this distance, other independent testing organizations have verified this capability. To accomplish this feat, Cylink uses a specially engineered radio and a high-gain 24 dBi semi-parabolic antenna system. This antenna is a bit bulkier than the Yagi-style antennas used by most of the other vendors. However, if your sites are 10 miles apart, it's good to know there's a wireless solution that will get the job done.

Unlike the other products, basic configuration parameters on the AirLink Bridge are handled via two banks of DIP switches on the back of the unit. While this approach is fairly effective for setting basic parameters, we found it somewhat limiting compared with the other products. For example, the unit's IP address, which is necessary for SNMP management, can only be assigned from a BOOTP server. We would have appreciated a serial-based terminal configuration option.

A feature unique to Cylink is its EIA-530 interface, which lets the bridge communicate via a wired interface to a leased data line. According to Cylink, some customers use its bridge products primarily on leased lines with an automatic failover to the wireless connection.

The AirLink bridge includes a series of useful LEDs on the front panel, including indicators for network activity, collisions, wire/radio mode and radio sync. Unfortunately, Cylink does not include a software utility to aid in aiming antennas for optimal performance and reliability. Instead, you must measure radio signal quality using a DC voltmeter attached to two signal-quality test-points.

The AirLink bridge is SNMP-manageable, including support for MIB II, a subset of the RFC 1286 Bridge MIB and proprietary extensions. While the implementation is not nearly as complete as that offered by Aironet, it will likely be adequate for most environments.

Cylink's documentation is the most technically detailed of the products. It includes detailed explanations of path analysis, fade margin, system gain and other radio-frequency topics. This focus reflects Cylink's strategy of distribution through value-added resellers.

The AirLink bridge's performance was lower than the other units. However, in fairness to Cylink, its full-duplex design, which was unique, would yield better results in a testing scenario that transmitted data in both directions simultaneously. If your site has bidirectional peer-to-peer traffic characteristics, you should factor Cylink's capabilities into your evaluation.

Dave Molta is director of network systems at Syracuse University. He can be reached at dmolta@nwc.com.

How We Tested Wireless Bridges

Testing wireless bridges is fraught with potential obstacles, both logistical and technical. Logistically, it is a challenge to test the products at their maximum rated distance, especially since the installations are temporary and therefore less optimal. For example, the antennas were mounted on tripods rather than secure antenna masts.

We began by selecting sites on the campus of Syracuse University, the site of one of our distributed test labs. For all but the Windata product, we tested the bridges in a line-of-sight environment with one antenna mounted on the roof of an eight-story building and the other secured to a tripod outside a building on a hill approximately 1.5 miles away. Since the Windata product had a distance limit of 0.9 miles, we tested that product between the window of our lab and the fourth-floor balcony of a building approximately one-third of a mile away.

We ran Novell's Perform3 benchmark utility between an Ethernet-equipped 486-33 laptop on one end of the connection and a 486-33 NetWare 3.12 server on the other end. In each case, the numbers were within 15 percent of those measured in the lab. Given our less-than-ideal antenna placements, we considered that sufficient evidence that these products perform as advertised over fairly long distances.

We were not able to verify the maximum vendor-rated distances for each product. Many options are available with respect to antennas, amplifiers and filters, so discuss your unique location requirements with a vendor prior to purchase.

In the lab, we tested the bridges for throughput using Novell's Perform3 utility, with four simultaneous stations connected to a 486-33 clone server running NetWare 3.12. For comparison, we included numbers for a non-bridged, direct Ethernet connection as well as a T1 connection between two Cisco 2500-series routers. The peak throughput numbers for Ethernet indicate that we were able to saturate our Ethernet test network to more than 85 percent of capacity. That indicates the lower performance of the wireless bridges can be attributed to true throughput rather than limitations of our test environment.

October 15, 1995