|First Flight - Amherst Takes to the Air (2.4GHz)|
© 2001 by Stephen A. Judycki
Amherst College was founded in 1821 as a non-sectarian institution for “the education of indigent young men of piety and talents for the Christian ministry.” Today, Amherst is an independent liberal arts college for men and women, and its approximately 1,650 students come from most of the 50 states and many foreign countries.
Residential and academic life at Amherst - as at many collegiate institutions its age – have undergone significant change because of technological advances. During my fifteen years at Amherst I have witnessed the latest of these advances, most of which were spurred on by the personal computer. As I began to think about the entire transformation – what it was like in 1821 compared to today - I quickly realized that the advent of public utilities was responsible for profound changes in the evolving “residential and academic experience.” I briefly describe some of Amherst's early physical culture to illustrate how those who came before us faced similar issues and probably dealt with them in similar ways. Our predecessors had to decide which new technologies to adopt and when to adopt them.
The contrast between early and recent advances is nothing short of striking. Wireless LANs on a campus once lit by lanterns and heated by firewood is evidence of Amherst's maturity. That her first flight with wireless LANs was not the first flight ever is evidence of her wisdom.
Witness to Invention and Innovation
The City of Boston, Mass., adopted gas lighting in 1822, followed by New York in 1823, but it was another 45 years before Amherst College considered gas for its buildings. In 1877 the Amherst Gas Company was formed with Amherst's president and treasurer serving as directors. [i] Amherst learned that piping gas into the 64 rooms and 16 halls of its two west dormitories was going to cost $612. This cost included fitting meters in each room, but not the cost of the meters themselves, “which the Company would probably furnish by charging rent for their use.” [ii] Amherst's buildings were fitted with gaslights from 1878 through 1885. [iii]
An 1849 alumnus described bathing in the early days of the Amherst College as involving an outdoor bathhouse that “was ten by twelve feet in size, and without a roof. A trough was built from the college well, and a student poured as many buckets of water from the well as he wished in his trough. The water flowed to a tank at the bathhouse. The student then crossed the grove to the bath house, disrobed, and pulled a cord, which released the water from the tank and provided him a cold shower.” [iv] Public water was introduced to the Town of Amherst in 1880. [v] Upon completion of renovations in 1891, water was piped throughout the first of two Amherst dormitories. [vi]
Stoves heated college buildings, and student rooms in the old dormitories were heated by fireplaces, both of which burned chestnut wood, until the 1890s when steam heat was introduced. [vii]
On October 12, 1901, the Amherst Student reported, “The electric lights on the campus were turned on for the first time last Friday evening.” Edison's carbon filament light bulb had won out over gas lighting years earlier, but being rural meant the arrival of most new innovations would be delayed.
In 1911 Amherst College signed a five-year, $242 per annum contract with New England Telephone and Telegraph for its first telephone system. The system consisted of a switchboard arranged for intercommunication between stations, an operator's station, two trunk circuits connecting the switchboard to the telephone company exchange, and 12 telephone instruments. [viii]
In 1926 the College's Committee on Buildings and Grounds issued a report stating, “It has been contemplated that the College would place in underground conduits the wires which in the past have been strung from building to building and to the trees on campus.” This did in fact happen, restoring the public appearance of the 150-acre main campus to its original, pre-utility character. [ix]
The Information Age
Let's fast-forward a half-century or so. While Amherst's mission does not require it to blaze a path on the leading edge of information technology, it doesn't lag very far behind. Amherst has been a wired campus since the mid-to-late 1980s. This, of course, ignores the administrative data system, which was installed on an IBM mainframe platform in the 1960s and continues to provide service today. Amherst College had installed a “port per pillow” long before that term became popular. Its first network platform was a proprietary CSMA/CD network that operated on a broadband coaxial system. For two or three years before converting to a fiber-based Ethernet network in 1995, Amherst used 10mb/s broadband Ethernet bridges to transport data across the campus. This technology was the predecessor to the cable modem service marketed today under such names as Roadrunner and Pipeline.
Ease of use and ease of access rank high among the challenges confronting the information technologist in higher education today. While Windows OS, Mac OS and the web browser have accelerated computer and network use by simplifying the user interface, the ease of accessing networked data has largely been a function of the proximity of the data jack to where someone wanted to work. In February 2000 Amherst's Director of Information Technology, Philip E. Fitz, was trying to read the wireless tea leaves. He knew that wireless LANs would remove a barrier to accessing information, and that could enhance both teaching and learning at Amherst. He wondered if the evolution of wireless standards and technology would keep pace with bandwidth-hungry applications, and whether wireless LANs could replace their wired counterparts in time to avoid the expense of another major wiring upgrade.
The installation of data wiring to the desktop – at least for us - was an evolving process. Computer resource centers and employee offices were wired first, followed by libraries and dormitory rooms, and finally classrooms and laboratories. Shortly after becoming a “wired” campus, we found that it was time to upgrade our wiring to accommodate faster data transmissions. It was costly but necessary. During building renovation projects we added many more data jacks. Sometimes jacks were added to existing wall plates, but more often new wall plates were added to new locations. As the number of data jacks increased, so did the number of hub ports and switch ports that were required to support them. Over time we had installed enough live data jacks for every member of our community to have somewhere between two and three network-attached devices plugged in and operating simultaneously. This, of course, is not a practical capability since many of the data jacks are located in public areas. But the end-result of all this wiring was the creation of an “anytime, anywhere” model of network access – a model that will be very costly to duplicate if and when wiring has to be upgraded to accommodate faster data transmissions once again.
The concept of connecting a computer to the LAN without a wire initially gives hope to the prospect that we may never have to install another data wire to the desktop. One of the more perplexing problems encountered in the installation of data wiring arises from the dynamic nature of the desktop – few desks are bolted to the floor, which makes them moving targets. Buildings under construction don't yet have any, which makes planning difficult. Old buildings that are renovated or re-purposed add new desks and move old ones. And people like to move their desks for various reasons - often to the opposite side of a doorway, which creates problems - most involve tripping. If wireless LANs are ready for prime time, they could put an end to this costly practice of wiring and re-wiring!
There is also, more importantly, a utility aspect of wireless LANs. Some people need to work, and others' work could benefit from working, away from the desktop. Wireless LANs enable mobility – the freedom to move around untethered and unrestricted. The first barrier to this kind of access has already been removed by the laptop and the PDA. In a library, for example, the laptop user can move about the stacks while accessing the online catalog. He can take notes while sitting in the periodicals section. She can ask questions at the reference desk while pointing to the item in question on her laptop screen. A student whose dormitory is being used as the scene of a rock concert would have new options for studying elsewhere. In the classroom, a faculty member who wants to supplement lecture with networked data would require less set-up time. Every classroom could become a computer classroom (student-provided of course) at the discretion of the faculty member. Clearly, wireless networking adds a new dimension to the anytime, anywhere model of network access.
OK. Let's Try Wireless
In early May 2000 I accepted the assignment of getting a wireless LAN pilot up and running by the start of the Fall 2000 semester. My technicians and I began with a sort of wireless brainstorming session that, in just a few minutes, produced a list of topics that included standards, security, management, interference, compatibility, interoperability, manufacturers, vendors, documentation and testing. This list provided a simple but effective framework for launching our project.
We began by surfing the web in search of useful information. My first stop was the IEEE web site, where I thought I would peruse the 802.11 standards. I didn't know that you had to buy them before you could read them, but it makes sense that even a non-profit organization requires income to cover expense – mine does. The Wireless Andrew pages of the CMU web site were very helpful with respect to design considerations. It was there that I first saw the term “airspace policy,” and realized its potential importance. As a GartnerGroup client, we read their reports and spoke with one of their wireless analysts. They provided planning assumptions that helped us decide between direct-sequence spread spectrum (DSSS) wireless LANs and frequency-hopping spread spectrum (FHSS) wireless LANs, both of which operate in the unlicensed 2.4GHz band. Other planning assumptions helped us to size up the interference threat posed by Bluetooth wireless LAN technology. Additionally, GartnerGroup identified which vendors were likely to be dominant in the future by providing measurements of their completeness of vision and ability to execute that vision. We also spoke with a handful ACUTA members who had already installed wireless LANs on their campuses.
One ACUTA member had an extensive fabric of wireless LANs, but couldn't recommend a vendor, because they ordered their wireless equipment directly and performed most of their design and installation themselves. As we weren't staffed to do much beyond post-installation maintenance, we educated ourselves in the area of radio design, equipment specifications and installation techniques, and then set out to find a vendor that knew more than we did. The first two vendors we tried did not know more, and even though they came highly recommended, I couldn't be convinced that they could do a good job. The third vendor we tried was Direct Network Services (DNS) of Ayer, Mass. In just five minutes of telephone conversation with Dan Kirkland of DNS, I knew I had found the right vendor for Amherst.
Our first meeting took place on June 5, 2000. Dan Kirkland spoke with extensive knowledge of industry standards, including the status of 802.11 subcommittees charged with developing the next standard, and he was well versed in market-share statistics. He explained that Lucent Orinoco (formerly Wavelan) had sold 80% of the client cards on the market worldwide, and added that all wireless access point manufacturers except Cisco Aironet (formerly Aironet) utilized cards made in the same factories off-shore, using the same chipsets. It was about this time that I realized that access points also use client cards, and the client cards for laptops and access points are interchangeable. I learned that the data rates available with 802.11(b) client cards are 11, 5.5, 2 and 1 mb/s. While early adopters of this wireless standard could only purchase 2 mb/s client cards, technology has advanced and today's cards operate at 11 mb/s. The standard provides for a fallback algorithm that throttles an 11mb/s client card down through these four data rates as distance from the card to the access point increases. Throughput remains at about 70% of signaling as the card falls back, but it's linear – loss is not greater at the lower signaling rates. Roaming between access points with a laptop utilizing an 802.11(b) client card is achievable when implemented properly, but this feature is not covered by the standard. Roaming is easily implemented with single vendor installations, but is sometimes impossible with mixed vendor installations. Several groups have been formed to ensure greater future compatibility and interoperability with such features as roaming, load balancing and bandwidth management. Among them are: WLANA –Wireless LAN Association, WECA – Wireless Ethernet Compatibility Alliance, WLIF – Wireless LAN Interoperability Forum and WIFI – Wireless Fidelity Standard.
The similarities between the DNS and GartnerGroup data were close enough that when DNS offered information in areas that lacked a GartnerGroup reality check, I perceived that information to be credible. This was especially helpful with predictions about the availability of 802.11(c) product and the likelihood of downward compatibility with 802.11(b) product.
Defining the Project
By mid-June, we decided that our pilot would involve the installation of wireless LANs in the libraries and public areas of five academic buildings – Robert Frost Library, Music Library, Merrill Science Center, Webster Center, and Keefe Campus Center. The chosen equipment would comply with the 802.11(b) standard and we clearly understood that this equipment might have a useful life of only 12–18 months. For the wireless LAN connections to be useful from the client perspective, we wanted the maximum 11mb/s data rate available in as much of the designated coverage areas as was possible. We didn't want coverage gaps within a coverage area, and we didn't want interference between overlapping coverage areas. We had design criteria that needed to be established somehow, so that results could be measured and documented. An RFP would do the trick, but a high level of contractual specificity can be a double-edged sword, and it tends to drive project costs up. I wrote the RFP, but I held onto it while I asked DNS to submit a proposal and statement of work. As it turned out, their proposal was only missing a few of the important points from my RFP, which they gladly added to their proposal. We were both happy with the document (copy below). It enabled us to purchase a wireless LAN design that did not lock us into a single product or vendor.
DNS completed the site surveys during the week of June 19, 2000. The following week, with the access point locations identified on the floor plans, we hired an electrical contractor to install data jacks for the access points.
We were leaning toward Lucent equipment, because of the GartnerGroup data and the validation we received from our peers, but Cabletron was the only 802.11(b) access-point manufacturer that did not require local collocated power for its access points. This was an important consideration for us since it would reduce installation time, cost and disruption. Cabletron access points use transformer-based DC power that is inserted onto the two unused pairs of the data cable that serves the access point. Since existing data wiring closets already have AC power, no additional expense is required to power the transformers. We considered having power insertion and extraction devices manufactured for us locally, to be used with Lucent access points, until we were advised that Underwriters Laboratories would require $36,000 to certify and “list” them. This cost far outweighed the cost of the additional electrical outlets we had sought to avoid, and it was nothing compared to the potential liability we might face from using an electrical device that did not carry the UL listing. We decided to go with Cabletron access points. Another feature that seemed unique to Cabletron access points was their simultaneous flash upgrade capability – we could upgrade all of our access points' software from a single location with a single command.
Cabletron access points support the Lucent-hybrid client cards that Apple was shipping with its laptops, as well as the 64-bit (WEP) Lucent client cards that we would be recommending and supporting for use with Dell laptops. Dell was shipping Aironet client cards with its laptops that were purchased with a wireless option. Because the Aironet client cards would not roam between two Cabletron (or Lucent) access points, we would have to advise our community to order Dell laptops without the wireless option, and advise them of procedures for having a Lucent cards installed locally. The access points themselves would be using 128-bit Cabletron client cards for the enhanced management features and the additional security they could provide.
On June 23, 2000 we signed a contract with DNS for the wireless equipment purchase and installation, which included acceptance testing and “as-built” drawings. As the electrical contractor installed data jacks, DNS followed behind them with the access point installations. During the site surveys and subsequent installations I observed the talent and experience that initially made DNS appear to be such an attractive partner. A few of my observations, shared here, may help to convey the value of a having a knowledgeable wireless contractor.
1. The lowest level of the Robert Frost Library – a 100' x 190' space with movable stacks throughout, and limited clearance between the stacks and the ceiling – presented a unique problem. I initially thought we would need four-to-six access points, and would receive a mix of poor coverage and no coverage, depending upon the position of the movable stacks relative to access points and the wireless laptop. DNS used the aluminum HVAC ductwork that traveled along the ceilings as a wave guide for signal propagation, and covered the entire area with only two access points. Furthermore, acceptance testing indicated that the weakest signal to be measured anywhere on the floor, even between and behind the book-laden metal stacks, was –80dBm. Looking back to the data rate chart in the DNS proposal, you will see that –80dBm falls well within the requirements for the 11 mb/s data rate.
2. On the 2 nd and 3 rd floors of the Robert Frost Library, banks of metal patron lockers were blocking the wireless signal, causing a total coverage gap in a fairly small area on their far sides. My expectation was for second access points to be dedicated to these small areas. DNS added omni-directional ceiling antennae to the access point client cards, and positioned them approximately 30' away on the other side of the lockers. The cost of an external antenna is roughly 15% of the cost of an additional access point.
3. In our Music Building, there is a bank of metal lockers built into an interior wall. In this case, the lockers were not obstructing the wireless signal, but presented an opportunity for DNS. They were used to reflect the wireless signal back into the space to cover an area that otherwise would have required another access point or an external antenna.
4. In our Merrill Science Library, there is a drop ceiling, and the plans called for two ceiling-mounted access points. When access point locations are identified on a site survey, they are usually considered to be approximate locations. The final location isn't determined until the installation is performed, and sometimes it is installed a slight distance away from where it was anticipated. DNS realized at Merrill, for the first time, the unique opportunity presented by ceiling tiles. They mounted the access point to the topside of a ceiling tile in the area identified during the site survey, rather than mounting it to the structural ceiling above it. This technique would enable them to simply move the tile to any other location in the drop-ceiling grid, if adjustments were required to provide optimal coverage.
By mid-July, the installation and testing had been completed. Every test point throughout the coverage areas was capable of achieving the maximum 11 mb/s data rate, which greatly exceeded my expectations. During the following week, access points were configured with IP addresses and connected to the network backbone. During this time, my network technicians also became familiar with the Cabletron management software. By mid-August, our Desktop Computing Services department, which would be responsible for supporting the wireless clients, had completed its testing of client card configurations in both Apple and Dell laptops. With the knowledge they gained, they developed end-user documentation and prepared letters, which announced the availability of wireless networking to both students and faculty. With a team effort, we met our goal of having a wireless LAN pilot up and running by the start of the fall semester.
The feedback we received from those faculty and students who participated in our wireless LAN pilot has been generally positive. There also were comments about how useful it would also be to have wireless LAN capability in faculty offices and dormitory rooms – our pilot project only covered public areas in five academic buildings. Participation in the pilot thus far, which is voluntary, has been somewhat less than what we anticipated. We believe the main reason for under-participation is the lack of a critical mass of locations. That is to say, we think people will be less apt to volunteer to use wireless LAN access until it becomes as ubiquitous as its wired counterpart. But, unfortunately, it's one of those chicken-and-egg situations. Ubiquity will be achieved by deploying wireless LANs throughout the campus, and we are committed to this task, but prudent management of financial and human resources requires us to move slowly. We will pick up the pace as soon as the IEEE formally adopts the next standard, and product manufactured to that standard becomes available. I still hold out the hope that wireless LAN data rates will one day meet or surpass those of their wired counterparts. When this happens, and as long wireless data security is at least as effective as its wired counterpart, the practice of installing wires and fiber within and between buildings will begin to wane.
In the short time that has passed since the implementation of Amherst's wireless pilot, a few things have changed. Cabletron has broken up into four separate companies. Cabletron is now the name of a holding company for the four new companies, one of which is Enterasys, which manufactures and sells wireless products. Lucent Technologies has spun off its enterprise networking to a new company called Avaya. Their wireless family of products is still called Orinoco. The successor to the 802.11(b) standard is going to be 802.11(g), which will define a >20Mbps wireless product that operates in the 2.4GHz range.
Steve Judycki is director of telecommunications at Amherst College. He is a member of ACUTA's Legislative and Regulatory Affairs Committee, and Chairman of the Cable TV Advisory Committee for the Town of Wilbraham, Massachusetts. He can be reached at email@example.com.
[i] King, Stanley. "The Consecrated Eminence" (Norwood, MA: Plimpton Press, 1952) 75.
[ii] Amherst College Library, Archives and Special Collections, Physical Plant Collection: July 6, 1868 letter from Prudential Committee to Trustees of Amherst College.
[iii] King 75-76.
[iv] King 25.
[v] Carpenter & Morehouse. THE HISTORY OF THE TOWN OF AMHERST, MASSACHUSETTS (Amherst, MA: Press of Carpenter and Morehouse, 1896) 423-424.
[vi] CATALOGUE OF AMHERST COLLEGE FOR THE YEAR 1891-1892 . (Amherst, MA: Published by the College, 1892) 51.
[vii] King 173.
[viii] Amherst College: May 11, 1911 contract between New England Telephone and Telegraph and Trustees of Amherst College.
[ix] Amherst College: January
13, 1926 report from Committee on Buildings and Grounds.