Linux in the Real World

Paul M. Sittler

Issue #12, April 1995

Leviathan: A Linux-Based Internet Information Server

Texas Agricultural Extension Service (TAEX) “helps people improve their lives through an educational process which uses scientific knowledge focused on issues and needs”. Much of TAEX's mission involves information transfer to the people of the State of Texas. Like many government agencies tasked with providing better service with a shrinking resource base, TAEX has long been interested in innovative approaches to better serving the public without the barriers of time and distance.

TAEX began experimenting with electronic information distribution in 1984 with dial-in bulletin boards. In 1992, we rescued an aging Compaq DeskPro 286 from the auction block. We transformed it into an experimental Gopher/PopMail/FTP (File Transfer Protocol) server. The surprising response to this soon filled the 40MB drive. We began an exhaustive search for a better platform.

The best server software ran on Unix platforms, but they were prohibitively expensive. We were already using some Unix platforms for mail and networking applications. We tried to use one as a gopher server, but the response time was inadequate. We realized that additional duty as an information server would require significant system upgrades. We would need to increase main memory, add more mass storage, upgrade the operating system, and obtain the separately marketed “software development system” (basically a C compiler, awk, and yacc). Unfortunately, any one of these items cost more than an entire high-end Intel-based PC! When we tried to get price quotes, we discovered that the TCP/IP networking package for the upgraded SVR4 operating system was “not yet available”. Coincidentally, our agency also received another decrease in funding levels.

About that same time, a Finnish college student added networking to his free “Linux” operating system. This free Unix clone ran on readily available Intel 80386/486 processors with inexpensive drives. Initial experimentation on an 80386-based box showed that it actually worked quite well! TCP/IP networking was thoughtfully included, as was a superb C compiler system. Linux was like no other Unix system we had ever used, as it combined some BSD features with some SVR4 features, while maintaining some POSIX compliance. Our initial confusion changed to the delightful perception that Linux provided a most sensible mix of desirable features.

We obtained a low-end 80486 machine, fitted with a network card, and installed Linux. Both UMNs gopherd server and NWU's GN combination Gopher/WWW server compiled easily, and “Leviathan” was born. From a user's viewpoint, Leviathan seemed to operate faster, and we hoped that it would be able to handle a bigger user load than the Compaq. We transplanted the information tree from the Compaq to Leviathan, and both machines ran side by side, tangible proof that simple (and obsolete) computers could still be useful.

The initial information served via gopher included the “Master Gardener” files, the TAEX personnel directory, and abstracts of all Extension bulletins and leaflets. Wherever possible, we also provided the full text of these publications. We put the TAEX Agricultural Software Catalog on line. The 200MB drive soon filled, and a 300MB drive was added. Leviathan began performing as a bootp and PopMail server as “extra duty”. We scanned several clip-art collections and made them accessible. Usage grew steadily. Soon, users from all over the world were logging in round the clock. The March 1994 access logs showed that 2,245 sites obtained almost 100MB of mostly clip-art at a rate of more than 500 accesses daily. Then, Mosaic happened.

We had been experimenting with the Mosaic WWW browser since December 1993, but it was under Linux and X-Windows that I first experienced an implementation of Mosaic that really impressed me. Mosaic under Linux was stable, flashy, and very useful. This caused me to review Mosaic for DOS/Windows platforms. While not as stable and full-featured as the Unix versions, we evaluated it as tolerable. The World Wide Web concept seemed tremendously important, so we began demonstrating Mosaic throughout the Agency in February 1994. Newer versions of GN had added the ability to serve a new protocol, Hyper Text Transfer Protocol (http), which was used to tie together a “World Wide Web” (WWW) of networked information servers. In May 1994, text pages were marked up for Hyper Text Markup Language (HTML), and GN began responding to http requests as well as gopher users. Leviathan broke the 1,000 accesses/10MB per day mark that month with 31,427 requests, 3,406 of which were from http (Mosaic) clients.

We added anonymous FTP access in August 1994 at the request of many users. Leviathan began distributing the National 4-H Enrollment Management software via gopher, http, and ftp access. In September, Leviathan also became TAEX's departmental CCSO qi/ph maintenance client. The 2,000 accesses per day watermark occurred in October 1994, when GN served 28,345 files (510MB) in 63,000 accesses, of which 60% (36,000) went to gopher clients. Access rates increased by approximately 10,000 per month in both November (72,300) and December (83,801). December 15th's 4,570 accesses broke the previous record of 4,370 accesses established the day before. December's average access rates were about 2,700 per day, with 20.6MB of files retrieved daily. Gopher type accesses still accounted for 51% of the accesses (42,371), but http accesses increased more than gopher accesses from the previous month.

Cumulatively, between February and the end of December, Leviathan served 188,672 files (3.71 gigabytes) to 33,542 unique machines in 434,025 separate transactions, of which 295,330 went to gopher users. During the same time, another 4,844 files (151MB) were retrieved via anonymous ftp. Leviathan was still serving as a bootp and PopMail server in its spare time, while accepting logins for maintenance of our part of the University-wide CCSO “ph” directory services database system. All this activity took place on an inexpensive 80486/33 computer sitting under a table, with neither monitor nor keyboard attached.

Information Server Content

Leviathan is a niche server, publishing information of interest mostly to adult distance learners and educators. We have the obligatory information describing the organization, and searchable abstracts and texts of selected agency publications. The clip-art collection has grown to over 1,600 individual images, from three states, available in multiple formats, with another 999 nearly ready. We added a small experimental slide image collection that may be accessed through a “Contact Sheet Image Selection” imagemap using a pointing device. The TAEX Computer Technology Group “OnLine” computer user newsletter is available online, and the “Master Gardener Problem Solver” has been extremely popular. The TAEX Software Catalog is available interactively and may be downloaded in PostScript, text, HTML, and WordPerfect formats.

Leviathan also serves documents from sources external to Extension. The Texas Telecommunications Strategic Plan and its Executive Summary are accessible. The National Performance Review documents are available. A large collection of Internet information, including all RFCs (Requests For Comment), FYIs (For Your Information), and STDs (STandards Documents) to date may be browsed. Leviathan includes convenient links to all WWW and gopher servers and enables easy access to other Cooperative Extension information servers. To aid others who wish to establish WWW sites, Leviathan provides a collection of icons (some developed locally) and links to pertinent tutorial and technical documents.

Future Directions

TAEX is preparing a newer, faster server with more storage capacity to better serve Leviathan's users. The immense popularity of the clip art collections has stimulated us to prepare even more. We are preparing an extensive collection of slide images that may be distributed freely for use in demonstrations and publications. We are updating our online personnel directory with plans to provide color photos and sound clips of everyone in the agency for easy access by the traditional media. The online personnel directory may someday be linked to a “home page” for every person that includes a description of individual expertise and experience. We are considering making several large topical databases accessible online. The popular Master Gardener series begs to be updated with color images, sound clips and animation sequences. We plan to add more on-line newsletters replete with color graphics. We are considering distributing some free software packages by gopher and http access.

Reaching Leviathan sidebar

TAEX is working with several other departments and agencies to stimulate the proliferation of more information servers. The South West Agricultural Meteorological Information Service (SWAMI) has been supplying pertinent weather information from a Sun workstation for several months. More recently, “Monarch”, the TAEX Planning, Performance, and Accountability Server, came on-line on a DOS/Windows platform to keep the public and state legislators appraised of agency activities. Three of the Texas Research and Extension Centers plan to establish information servers. Two more experimental servers using alternative free software are currently “under construction” as TAEX explores this evolving technology.

The combination of telecommunications and computer innovations will together produce a technological imperative for change. This may require a major paradigm shift from information distribution toward providing information access. TAEX is preparing for the day when all of the information produced and distributed by the agency may also be made available digitally and online. With luck, it may actually be ready when the public is.

Impact of The TAEX Information Server

Leviathan has achieved some degree of national and international recognition. The NCSU list of “Top Ten Home Pages by Cooperative Extension Workers” points to us, as does their list of “Top Ten Extension-Related WWW Pages”. Other Cooperative Extension services across the US and in Australia, Mexico, India, Israel, and Czechoslovakia have downloaded clipart. A cookbook, published in the UK by an author from Bangladesh, incorporated some of the clip art as illustrations. An Australian server began mirroring Leviathan down under. Leviathan has been mentioned in several Canadian Agricultural publications. Leviathan was designated as a “Gopher Jewel” by several sites, and America OnLine lists it as their first agricultural information site.

The on-line software and image distribution has cut distribution costs considerably and made TAEX products more easily accessible. For example, Leviathan distributed 2.6GB of clip art on-line, saving taxpayers almost $17,000, compared to traditional floppy-based distribution methods. Each download of the Software catalog saves $5.00 in direct printing and mailing costs and has increased software distribution activity substantially. The National 4-H Enrollment Management software, distributed across the Internet has saved taxpayers more than $1,200 so far and provides more timely dissemination of updated versions. The 500MB of “Master Gardener” files would have filled 100,000 pages if printed, but being available electronically has saved $4,000 to date in printing costs alone. Most importantly, Leviathan has created an awareness of information server technology as a viable adjunct to traditional information distribution techniques.

Usage Analysis

While TAEX has received many favorable comments via e-mail, a more accurate progress assessment can be made by analyzing user access patterns.

Between February 13 and December 31, 1994, certain trends emerged. Daily usage patterns were high between 9:00 AM and 5:00 PM, but there were no “dead” hours. Weekend usage was about half that of weekday usage. Mosaic (http) access grew rapidly, but represented only 49% of Leviathan's total accesses. Gopher access was still important in 1994. The majority (62%) of the users were in the United States. Of these, most (67%) were from educational institutions, 16% were from the commercial (.com) domain, and 6% were from the government (.gov) domain. Leviathan's users were mostly (80%) from outside the TAMU system. Some 69,476 accesses (16%) were from Texas, which means that 84% of the requests were from out of state. Texas County Extension Agents visited 1062 times, while 215 Texas Extension Specialists connected 38,561 times.

Leviathan has been visited by people in 62 identifiable countries outside the United States on six continents. There have been contacts from 793 different educational domains in all 50 states. Delphi users accessed Leviathan 3,365 times, and America OnLine subscribers called 2,172 times. Compuserve users finally connected 48 times in December. The server has been contacted by 1,270 individual host machines more than 50 times; 561 of these visited more than 100 times; and 22 dropped in more than 1000 times in that period.

Lessons Learned from the Leviathan Experience

  • If you build it, they will come....The demand for on-line information is staggering.

  • If you provide useful information, they will come back, repeatedly, for more.

  • Providing networked access to information is often cheaper than traditional methods of distribution.

  • Networked access to information is only useful as to those who are networked. It should be viewed an adjunct to existing distribution methods.

  • Simple text-based information is valuable. FTP and gopher are not dead. While the sizzle may sell the steak, content is more important than presentation.

  • Consider your audience. Ensure that documents make sense when viewed in a text-only mode.

  • Use graphics sparingly. Keep in-lined graphics small. Many users have slow dial-up network connections.

  • The Internet does not end at the state line. We now serve people in other states and nations of the global village at no incremental cost.

  • Servers must serve 24 hours a day. It is 3 PM some where on the Internet all the time.

  • Servers should not be shut down for update and maintenance. This implies a multitasking operation.

  • Graphics and audio files should be clearly identified through use of icons.

  • Keep menu pages small and simple, with 5 ± 2 selections.

  • Almost any old obsolete computer can be used as a server. Storage capacity is more important than powerful CPUs.

  • Current WYSIWYG editors are often harder to use than simple text editors for producing HTML.

  • Storage space and network bandwidth are both finite.

  • Time costs more than equipment.

  • Administering a server takes more time than expected.

  • Organizing information is sometimes harder than producing it.

  • Writing HTML documents is far easier than it first appears.

  • We often overestimate what we can do in a week and underestimate what we can do in a year.

  • Linux is not merely a hobbyist's toy; it is a solid, stable, professionally-implemented Unix clone that performs superbly as a production information server platform.

Summary

Evolving telecommunications and computer technology are combining to produce a technological imperative for change, requiring information access as well as traditional information distribution techniques. Government agencies will continue to be asked to provide more and better services with a shrinking resource base. Demand for information access will likely continue to increase for the foreseeable future, causing a corresponding demand for more network bandwidth. The TAEX combination FTP/Gopher/WWW server demonstrates that a low cost (> $1,500 US) computer running only free software can supply a moderate-to-large amount of information. Usage is growing at a rate of 10,000 accesses monthly. Demand for gopher http access. FTP is still a viable and desired method of file transfer. Information servers provide information to inhabitants of the global village without barriers of time and distance. TAEX's Gopher/WWW/FTP information server is a promising alternative medium for outreach and distance learning for our changing clientele. It is a digital extension of the agency's motto: People Helping People.

Paul M. Sittler (p-sittler@tamu.edu) is a Computer Systems Engineer for the Texas Agricultural Extension Service. He enjoys playing with technology and making it useful to others of his species. He received a BS and MS in Vocational/Industrial Technical Education from Texas A&M University.