|Bridge Builders: African Experiences with Information & Communication (BOSTID, 1996, 304 p.)|
|Case studies on electronic networking|
bv Paulos Nyirenda
Dr: Paulos Nyirenda is Head of the Departnent of Physics and Electronics at the University of Malawi. He has a PhD in electrical engineering from the University of New South Wales. He has been working on networking at the University of Malawi since 1992. He asks that this chapter be dedicated to his late wife, Gemma.
BACKGROUND AND CONTEXT OF THE PROJECT
Malawi is a landlocked country located on the southeast side of Africa. It shares boundaries with Mozambique, Tanzania, and Zambia. The country has a population of about 9 million, of which about 90 percent live off subsistence farming. Malawi's economy is agricultural-based, with few manufacturing or mining industries.
The University of Malawi is the only university in the country. It is made up of five colleges spread out over a wide geographical area in the central and southern sections of the country. There are many research institutions in the country, most of which are linked to the agricultural sector and, as such, are located in remote locations across the country. Most of these research centers have access to direct exchange telephones that work most of the time but are of relatively poor quality, just as in some other African countries. Also, as in other African countries, the telephone penetration into the population in Malawi is very low.
Communications in Malawi is difficult and expensive. Research and data communications have been achieved primarily by physical travel to a site - often over seasonal roads in poor condition - or by fax where available. This is also the case for most governmental and private sector communications. Most needed research and other data and information do not reach the people and decision makers who require this data to make important national and international decisions. However, even as early as 1992, most research, governmental, and major commercial offices had computers. Most of these were IBM-compatible desktop computers or Apple Macintosh computers.
Realizing the difficulty of communicating among widely separated university colleges and research institutions, I proposed a project titled Study of Computer and Telephone Network Based Communications in Malawi. The proposal was submitted for funding to the University of Malawi (UNIMA) Research and Publications Committee (RPC) in February, 1992. The original budget was for Malawi Kwacha (MK) 5123' or roughly $1,000. The main objectives of the project were to:
· investigate the feasibility of establishing computer and telephone-based communications in Malawi; and· demonstrate more efficient and effective communications among researchers and academics within Malawi as well as among those in Malawi and outside.
With these objectives and minimal base funding, the Malawi Fidonet network was started. The network, now called UNIMA, has grown to be a public, nationwide network serving all sectors - government, non-governmental, and commercial - of the Malawi economy. It thus serves a much wider population than the academic and research communities originally envisioned and for which I had budgeted. It is a public network in the sense that anyone and any organization in Malawi can be linked to the network - provided they have the basic resources required for such linkage. The project has since attracted funding from various national and international organizations.
PROJECT EXPERIENCE AND IMPLEMENTATION
The proposal I made to the University led to the establishment of the UNIMA Fidonet network. The specific objectives of the proposal were to:
· experiment with modems, microcomputers, and the telephone network for data communications within the University of Malawi;· design and build electronic interface devices for connecting direct telephone lines and switchboard (PABX) extensions to one computer and modem node to handle both local college and direct line connections;· evaluate the quality of the Malawi telephone network for data communications, and · evaluate the cost of computer communications using modems and the telephone network and compare it to other electronic communications modalities.
At the time the UNIMA project was started, there was a regional project called the East and Southern Africa Network (ESANET) funded by the International Research Development Centre (IDRC) of Canada. (For a complete description of the ESANET project, see Charles Musisi's case study on page 158.) The UNIMA project had similar general objectives to ESANET, which linked university computer centers in Zimbabwe, Zambia, Tanzania, Uganda, and Kenya. Because I was involved in these activities, the UNIMA project benefited considerably from the ESANET project in the beginning, even though Malawi was not a participating country in ESANET. The startup modems and non-commercial Fidonet software that enabled UNIMA to take off were obtained under ESANET.
My proposal to study computer and telephone-based communications in Malawi was accepted by the RPC in January 1993. I next needed to request security clearance from the Malawi Government, as well as from the Malawi Posts and Telecommunications to start the project. The government issued the security clearance on 15 June 1993, one day after citizens passed the referendum that introduced multiparty politics into Malawi. The Malawi Posts and Telecommunications Department then cleared the project in July 1993. The direct telephone line required for the node or hub of the telephone-based computer communications network was installed at Chancellor College in Zomba on 24 September 1993.
We next installed the equipment for the node: an IBM-compatible 386 computer running at 20 MHz with a disk space of 40 megabytes (borrowed from the Physics Department at Chancellor College); a VIVA 2400 baud modem donated by IDRC under the ESANET project; and non-commercial Fidonet communications software (FrontDoor 2.02) obtained under ESANET and tested earlier. This startup setup was then used to negotiate the network address, initial mail routing, and initial polling procedures required for the network. In the startup phase, the network address was negotiated under ESANET to be a Fidonet point address off the University of Zambia electronic mail network. Malawi now has its own network number and the node established has the Fidonet address 5:7231/1.(2)
Connecting to the University of Zambia Fidonet node was very difficult and, when the connection was finally made, the line quality was poor. We decided in November 1993, to switch connections to the Africa Zone Gate directly at Rhodes University in South Africa. This improved the connection success rate as well as the quality of the connections. In May 1994, the UNIMA network started receiving funding from the Capacity Building for Electronic Communications in Africa (CABECA) project also funded by IDRC but managed under the Pan African Development Information System (PADIS) in Addis Ababa, Ethiopia. (For more information about the CABECA project, see Fall's case study on page 147.)
Under CABECA, the UNIMA network received more modems and assistance from the Southern Africa Non-Governmental Network (SANGONET) in Johannesburg. SANGONET polled the UNIMA system twice a day at first but this was upgraded to three times a day to cater to the increased volume. Also under CABECA, we upgraded the node with the installation of a Telebit WorldBlazer high speed modem. This modem was eventually replaced by a US Robotics modem.
Figure 1 shows the general Malawi Fidonet network and its linkage to other networks in the world. The number of points running off the node has increased rapidly even though the single most important constraint to the growth in the user base has been the availability of modems at an affordable price within Malawi.
Figure 1: Fidonet structure in Malawi
We replaced the computer borrowed from the Physics Department with a dedicated computer borrowed from UNIMA RPC. This was also an IBM compatible 386 computer with the same speed but with a 104 megabyte hard disk and more memory. This computer made it possible to perform multitasking using DesqView and QEMM. This allowed us to operate two telephone lines, using intermail as the communications program on the node. With this setup the node is now able to handle two data connections simultaneously, one on a direct exchange telephone line and another on a PABX telephone extension that caters to the local Chancellor College campus computers. Putting in the line from the local PABX helped to reduce congestion considerably on the direct line.
The use of a multiline mailer that enabled the node to handle more than one telephone connection at the same time made it unnecessary to develop hardware devices to achieve the same results. However, there were problems with rescanning the mail every time it was collected from the node. As the message base and user base grew, this mail rescanning took more and more time and eventually led to lengthy in-between call processing, leading to delays in answering calls at the node.
After observing the significant contribution that the Fidonet UNIMA network was making towards improving data communications in Malawi, the World University Service of Canada (WUSC) provided funds to the UNIMA network in February 1995 to purchase a Gateway 2000 IBM compatible 486 DX2 computer with 730 megabyte hard disk and running at 66 MHz. This has considerably improved the performance of the node, which can now process mail much faster and has removed the waiting that users experienced between calls while the node computer was processing received mail and extracting billing information. In addition the network has also received support from the Canada Fund of the Canadian High Commissioner to Malawi to purchase modems to assist the public sector in getting connected to the network so as to improve their communications.
UNIMA Network Operation
Under Fidonet, electronic messages and files are prepared offline to reduce telephone connection costs. The messages and computer files are then transferred onto the network in compressed files or packets to reduce the time of the telephone connections and to improve efficiency. When a point has a message to send, the message is normally sent to the node to be routed to its destination. If the message is destined for another point on the UNIMA network, the message waits at the node to be picked up by the destination point. Before April 1994, the node originated international calls to the Africa Zone Gate at Rhodes University for delivery and collection of international mail. All international mail is now routed via SANGONET, which polls the UNIMA node three times a day. International mail from Malawi is made to wait at the Malawi node in Zomba for international delivery during one of the polls from SANGONET.
When mail is received at the UNIMA node, it is processed by the non-commercial Fidonet mailer and other mail processors. Gecho.exe (version 1.01) is used as a mail and conference processor and also for mail compression and decompression. Netmanager, Netmgr.exe (version 0.99), is used as a general message processor for message-by-message identification, distribution, and redirection. Message tracking is done by msgtrack.exe, which produces a message-by-message log indicating such data as message origin, destination, dates, and volume.
I have written additional software to produce bills sent to users for cost recovery and network sustainability. Most of the billing information is obtained from the data produced by msgtrack.exe. Other data used to evaluate the performance of the network and the telephone system is collected from the various log files produced by the mailer and mail processors and analyzed using tools developed at the UNIMA node.
RESULTS, IMPACT, AND BENEFITS OF THE PROJECT
The network performance reported here was monitored at the UNIMA Fidonet node. As outlined above, the node software produces log-files (logs) for the day-to-day, telephone call-by-call, as well as the message-by-message activities that the node performs in sending and receiving electronic mail and files. The collection of the data presented here started in October 1993 and ended in March 1994. Much more data has been collected at the node but this has yet to be analyzed. By processing the log files, the monthly international mail volume can be determined. This is shown in Figure 2.
Figure 2: International e-mail volume and telephone bill using a 2400 baud modem.
Figure 3: Cost per kilobyte of international telephone calls for e-mail transfer using a 2400 baud modem.
Cost of the Service
Shown in Figure 3 are the costs of sending or receiving one kilobyte volume of mail and the monthly telephone bill in Malawi kwacha, as charged by Malawi Posts and Telecommunications Department. This calculation was done on a month-to-month basis as well as on a cumulative basis where the total bill and the total volume, up to the dates shown, were used to calculate the cost per kilobyte of mail. As shown in Figure 3, the cost of operating the network per kilobyte of mail has dropped from MK4.68 (four Malawi kwacha and sixty-eight Malawi tambala) in October 1993 to MK0.88 in March 1994 on a month-to-month basis and from MK4.68 to MK1.10 on the cumulative basis. Using these data, the node has recommended that users be charged at the rate of MK1.00 per kilobyte of mail sent or received internationally. This charge rate was still in place at the writing of this chapter.
Although the modem used for collecting the data changed from a 2400 baud modem to a 14400 baud one (with most good connections at 9600), the Malawi telephone charge per minute also changed upwards considerably from about $1.00 to $3.00 per minute in January 1995. We are currently collecting and analyzing data to check the effect of the modem speed change versus the telephone charge rate change.
A comparison between the cost of using the email system and using a fax to transfer text messages can be made based on the above data. One full page of text (as on this page) uses nearly 3 kilobytes and would therefore take MK3.00 to send by email. After examining several fax transmissions at Chancellor College, we found that such a page would take more than one minute to send but often less than two minutes. (The Malawi Posts and Telecommunications would therefore charge two minutes.) Most of the international mail transferred during the observation period has been for contacts outside Africa. The cost of sending such text messages internationally by email is therefore about 20 times cheaper than faxing similar messages at the current telephone rates and email charges at the UNIMA node.
Using the data collected from message tracking, a bill is sent from the Bursar of Chancellor College in the University of Malawi to each user once every three months. In addition to the volume fee outlined above, which now stands at roughly seven cents per kilobyte, each user account held outside the university is charged at 70 cents per month for account maintenance. Experience has shown that although these rates are very low, most Malawian users and organizations still have difficulties paying. I am currently proposing to conduct a study to find out whether this is due to genuine financial difficulties. I hope that user perception of "user pays" electronic communications will be better understood following the survey and analysis of the results.
I have shown in Figures 2 and 3 that even though the monthly bill and the monthly volume of mail sent and received internationally are increasing rapidly from month to month, the cost of sending mail per kilobyte has steadily approached a constant value. I expect that the speed of the modem used at the node will play a crucial role in determining the cost per kilobyte of international mail transferred at the node. At the moment this has been complicated by the international charge rate change made in January 1995. I hope that the further data analysis now being carried out will assist in clarifying the situation.
All the money raised from the bills is paid into a project account held at the Chancellor College Bursar office. Money in project accounts does not get absorbed into the University's pool account; although the money does not generate interest in such accounts, it is available when required to pay the bills incurred by the project. It also helps cover equipment maintenance and staff costs.
Figure 4 shows the failure rate of telephone calls for data transfer as monitored at the node. All the calls at the node involve the Malawi Telephone system. Data transmitted during a call that fails in some cases have to be re-transmitted because the decompression of the data at the receiving side fails due to incomplete packets or files. Thus a failed call is a major concern to a user on the network. As can be seen from Figure 4, the proportion of calls that fail has come down considerably from nearly 100 percent in October 1993 to about 18 percent in March 1994. This improvement in performance can be attributed to various factors. The international mail route changed in November 1993 from via Zambia to via South Africa. We observed that fewer calls fail on the new route. We have also learned when to make the international calls in order to achieve higher success rates. This information was not available during the startup phase of the network.
Figure 4: Failure rate of telephone connections using a 2400 baud modem.
The number of calls to the node from Fidonet points within Malawi has grown considerably as a proportion of the total number of calls recorded by fdstats.exe at the node. This means that while the performance reported in Figure 4 applied more to international calls in the startup phase of the network, it applies both to national and international calls in March 1994. We can estimate from Figure 4 that on the Malawi telephone system about 18 percent of the calls for data transfer using modems will fail. More data needs to be collected and analyzed to determine the separate failure rates of national and international calls on the Malawi telephone system and to determine the rate which the graph in Figure 4 settles down to.
ANALYSIS OF LESSONS LEARNED
The main services on the Malawi Fidonet network are electronic mail and file transfer. No remote login of any kind is currently supported. We need to conduct a survey to determine and categorize the purposes of sending mail by users within the University of Malawi and outside. During the experimental phase of the network, we have sought such information from users in the University and outside. The message tracking software at the node collects information on the sources and destinations of mail. We have observed that University of Malawi users use the Fidonet network mostly for the following reasons:
· contacts with academic colleagues and supervisors;· negotiating links with other universities at department level;· submission of research papers, proposals and reports;· applying for further studies abroad;· coordinating research for staff on study leave abroad'· library contacts;· seeking funding for research and other projects;· discussion of issues pertinent to Malawi;· ordering equipment from suppliers outside Malawi'· personal mail; and· file transfer using archie and ftp by email.
In using the network, University staff and other users are finding it easier and more efficient than before to request and access information from colleagues and institutions outside Malawi. I hope that as the network grows inside Malawi it will also become easier and more efficient to access colleagues and information sources within the country.
A number of users on UNIMA were already on email in Malawi but they used to make international calls to access email facilities abroad, such as CompuServe or CGNET. Since the UNIMA Fidonet network was established, many of these users have switched to using the UNIMA network and are finding it cheaper and more convenient since it only involves a local telephone call. (See Box 1.)
Toward the end of 1994 and the beginning of 1995, users in non-governmental organizations and commercial companies began making up a fast-growing proportion of the user base. These are high volume users who request and send large amounts of data over the network - mostly involving business related information. I expect that in the near future these users will be the main revenue generators for the running of the network.
Growing the Internet
With more than one hundred points on the UNIMA Fidonet network and over six hundred users, it is time for the establishment of a full Internet link in Malawi. The user base is there to support the network and our experience has shown that the users would also be able to pay the fees that such an installation would demand. At the point of writing, a leased line to South Africa seems too expensive and so we are suggesting that a VSAT connection be tried instead. VSATs or very small aperture terminals are satellites that provide two-way, high data rate services. VSATs appear to offer a low-cost telecommunication solution in developing countries.
Users on the UNIMA network indicate that they would not like flat billing. They would prefer message-by-message billing so that they pay for what they use. We will, therefore, have to carefully examine the current billing practices on UNIMA before full Internet connectivity is installed.
The main constraint on the expansion of networks in Malawi will remain equipment based, with modems topping the list. I recommend to donors that, when projects are funded in a country like Malawi, each project should have an information technology section providing for the supply of computers and modems, when and where these are not available. Donors should also consider the expense of the installation and training of users of the information technology within the project. The minimum here should include electronic mail in any project proposal and implementation.
BOX 1 Reviewing the New Constitution
The UNIMA network made a significant contribution in the development of the new Malawi constitution. It made it possible for international lawyers in Malawi and outside to exchange information in a fast and more efficient way. It promoted discussion and editing as the constitution was sent worldwide for comments and review. This was reported in the local press in Malawi ( The Nation). It was said that, without this mode of communication, the review of the constitution would have been lengthy and it would not have been possible to meet the deadlines for the review.
CONCLUSIONS AND RECOMMENDATIONS
A project to study computer and telephone network based communications in Malawi has been in progress in the University of Malawi since 1993. Under this project a nationwide electronic network using Fidonet technology has been installed and is being expanded. The network supports electronic mail and file transfer. This is a public network in the sense that anyone or organization with the required resources can link into the network. Most of the objectives of the project have been accomplished.
The project has shown that using the installed network for electronic communications is cheaper and more convenient than existing methods, such as fax and voice telephone. It has also shown that the Fidonet software can be used to generate data on a local telephone system that can be used to monitor the performance of the local telephone network. We expect that such data will be useful to the telecommunications operating corporation.
We have had many positive and congratulatory comments on the network performance and about how it has improved the communication efficiency of our users. So while the project began as a scholarly undertaking, mostly to determine the feasibility of establishing computer and telephone-based communications at the University of Malawi, it has resulted in the provision of improved communication services throughout the country. The small amount of seed money that we received from the University has been effectively multiplied and the impact of the project has been felt far beyond what we had originally intended.
1. The value of the Malawi kwacha has changed considerably from roughly US$ 1 =MK5 at the time when this date was collected to US$1=MK15 roughly et the time of writing of this report.
2. Fidonet uses a hierarchy to pass messages through the system. A Fidonet address identifies the zone. the hub, the node, and the point (if there is one) number of the user. This address, 5:7231/1, means Zone 5 (Africa), hub 7231, and node number 1. See Fall's case study for a more complete discussion of Fidonet.