April 30, 1995
Attila Özgit <ozgit@metu.edu.tr>
Kürsat Çagiltay <kursat@metu.edu.tr>
Erdal Taner <erdal@metu.edu.tr>
Abstract
This paper describes recent networking activities in Turkey, along with some common problems of organization and funding that most of the developing countries are facing today. We believe that some of our experiences could be instructive for similar cases in other countries.
Wide Area Networks in Turkey were first organized in 1986. First international network connection was made to EARN through a 9600 bps leased line to France. Turkey's Internet connection was established in April 1993 by a 64 Kbps leased line to NSFnet (Washington, USA). Since then, activities of TR-NET project group in the last couple of years have placed Turkey to the list of fastest growing IP domains.
Turkey's international traffic through NSFnet has increased by more than four-fold within last 18 months. As TR-NET has become wide-spread with a tremendous rate of growth among various sectors (academic, government, private sector, etc.) in 1993 and 1994, a severe necessity of governmental support and constitutional framework is being observed. Having anticipated the need for central planning and implementation, the TR-NET executive committee has submitted a proposal to the government for becoming an official coordination point for establishing and diffusing computer networking infrastructure in the country. This new organization is about to be established and will be known as 'National Computer Networks Center (NCNC)'. The NCNC will have the responsibility of laying down strategic plans and controlling the implementation for developing the national computer networking infrastructure.
Since currently available funds of TR-NET are not sufficient to constitute a national backbone and to enhance the international connection, TR-NET administration is currently looking for some funding models to overcome these problems. Some alternative models for funding have been investigated and are still being evaluated.
One basic model approach is the privatization of access services through multiple service-providers. In this approach, the backbone will still be operated by the TR-NET organization, while user access to Internet will be provided by private service providing organizations (SPOs). In this model of organization, the country-wide backbone and international connections will be funded from two different sources: (i) funds allocated by the government, and (ii) the service charges that would be collected from service providing organizations (SPOs). It is also anticipated that some other variations of the above model could be constructed. In this paper, a hybrid model of funding and organization is discussed.
In order for solving some of these problems, during the years 1991 and 1992, a project was initiated by the Middle East Technical University (METU) and the Turkish Scientific and Technical Research Council (TUBITAK) for Turkey's Internet connection [2]. The first step was the connection to RIPE in October 1992 by using IP over X.25. Turkey has established her dedicated Internet connection to NSF by the beginning of 1993 [3]. Since then, Turkey's international network traffic has been exponentially growing. By the beginning of 1994, Turkey's EARN gateway connection from Ege University to Montpellier, France was diverted to Bonn, Germany with an increase in bandwidth (64Kbps), and the line protocol was changed to allow IP traffic.
In parallel to establishing Turkey's Internet connection in 1993, METU and TUBITAK have also formed an informal organization known as TR-NET, to promote the use of Internet technologies in various sectors of the country. As a result of these efforts of TR-NET, today, government, military, commercial, educational and research sectors are actively using these network facilities established and offered by TR-NET. Personal connections and utilization are also increasing.
Today, the proposal for a well-developed TR-NET organization to become the official center for national computer networks is about to be completed (i.e., to be accepted) by TUBITAK. The model for funding such a country-wide network is also being discussed by the TR-NET executive committee.
Section-2 of this paper includes a discussion of current networking infrastructure and basic issues on development of a national networking infrastructure for Turkey. In section-3, future development plans along with a list of actions to be taken are discussed. Section-4 discusses organizational issues and funding model for a national networking initiative. Finally, section-5 gives a summary of the paper and specifies those major steps to be taken next.
TR-NET services are being centrally provided from Ankara, by METU and TUBITAK. Two SPOs, Istanbul Technical University (ITU) and Marmara Research Center (MAM), are extending TR-NET services to Istanbul region. However, users' demand for being connected to TR-NET is constantly increasing. Only in the first three months of 1995, there were nearly 600 personal applications.
Leased-line connections of TR-NET are shown in Figure-1. This basic infrastructure of TR-NET also overlaps with the infrastructure of TUVAKA that is also discussed in the following paragraphs.
Figure 1. Leased-Line Connections of TR-NET.
TR-NET offers other types of connections, such as, X.25 and dial-up. X.25 and dial-up connections are shown in Figure-2. Each line in the figure represents several connections from a specific sector (government, private, etc.).
Institutional and individual connections for using TR-NET services are discussed later in this section.

Figure 2. X.25 Connections of TR-NET
TUVAKA's (Turkish Universities and Research Institutions Network) current topology that has evolved during the years from 1986 to 1994, includes 29 of the 52 universities in the country. Those universities that were established last year are being connected to TUVAKA in 1995. Current topology (as of 03/95) is shown in Figure-3.

Figure 3. Topology of TUVAKA
Most of the institutional connections within the country are of 19.2-28.8 Kbps leased or X.25 lines. Personal or institutional connections using dial-up technologies mostly use 14.4 and 28.8 Kbps telephone lines. Backup speeds for leased lines, such as 9.6 or 2.4 Kbps are also used as required.
Following sub-sections discuss those issues that are considered to be the most important facts and/or problems of today's networking in Turkey.
Universities' and research institutions' network connections are centrally funded by TUVAKA. Nevertheless, the total amounts of funds are not sufficient for setting up a well-developed infrastructure for academic institutions. Annual budget of TUVAKA is about 400,000 US Dollars that is mainly used for only paying PTT line costs of currently connected universities. There is no centrally available budget for investment (i.e., procurement of equipment, staff, other initiatives, etc.).
Having anticipated the need for central planning and implementation, TR-NET organization has submitted a proposal to the government for becoming an official coordination point for establishing and diffusing computer networking infrastructure in the country. This new organization is about to be established and will be known as 'National Computer Networks Center' (NCNC) [4].
The NCNC will have the responsibility of laying down strategic plans and controlling the implementation for developing the national computer networking infrastructure.
Forming such a central official organization will bring the following benefits: (i) central, thus optimal fund management for national resources, (ii) standardization on technology and operation, (iii) planned investment on technology and know-how, and (iv) central technical and operational support for the planned country-wide backbone.

Figure 4. Turkey's International Traffic Pattern
The figure reveals the fact that in last 6 months (i.e., 10/94-03/95), traffic curve has been flattened. This is due the saturation of the available international bandwidth that is currently 64 Kbps. Total host count is nearly 3,000 and the total number of daily users are estimated to be around 10,000 to 15,000. The demand for network bandwidth is constantly increasing, and the delay characteristics are getting worse.
Total number of individuals (personal use) has increased from a few hundred to 1300 within 16 months. Only in the first three months of 1995, there were 580 personal applications to TR-NET for connection. With the current level of service provided by TR-NET, and the rate of growth on user demand, it is estimated that there will be 3000 individuals connected to TR-NET by the end of 1995. Providing good quality service to thousands of individuals is not an easy task to handle. In order for handling this problem properly, and for providing individuals and institutions (government or private sector enterprises) always with top quality service, another goal has been set. This other short-term goal is the privatization of connection services.
Privatization will help in provisioning and fast diffusion of networking services with a much better quality of service to the end-users.
TR-NET is currently planning to identify candidate Service Providing Organizations (SPOs) in three major cities, namely Istanbul, Ankara, and Izmir. These three major cities cover almost 25% of the whole population and most of the commerce, industry, and government sectors. Each SPO is expected to be capable of handling at least a few hundred users by providing sufficient level of computing and networking resources as well as skilled service staff.
Forming such a central organization (NCNC) for strategic planning and management of the country's backbone network and extending services to other major cities will bring the following global benefits to the attention of Turkey's networking community:
The country-wide backbone that is being planned for the third quarter of 1995 is depicted in Figure-5.

Figure 5. Planned Backbone
Establishment of such a backbone network including only the three major cities of the country basically requires associated funding for equipment and lines to be granted. Total operational budget for maintaining a triangle country-wide backbone and the international connection is estimated to be around 1,000,000 US Dollars per year. Since the currently allocated governmental funds of TR-NET are not sufficient for such a backbone, one viable alternative would be to make use of funds to be obtained from privatization of services.
The other important long-term plan is the improvement of both quality and quantity of user services through a privatization and competition process. It is foreseen that there will be a lead time for such private services to be stabilized in terms of quality and price. However, it is our belief that the lead time will be minimal, and stabilization of services will be completed before the end of 1995.
The most important part of this new organizational model is the introduction of service providing organizations (SPOs) layer between TR-NET backbone and the users. SPOs are qualified as private companies, specialized in networking and information technologies. They will provide the user community with information, consultancy and connectivity services as well as provisioning of required hardware and software components. SPOs are expected to market all their assets as value-added services on top of the basic backbone connectivity and information services of NCNC.

Figure 6. Organizational Model
The inner-most layer is the country-wide backbone. This backbone will be maintained by NCNC, and will initially cover three major cities, namely, Ankara, Istanbul, and Izmir. NCNC is planning to offer contracts to eligible SPOs for extending services to the user community. Organization of NCNC is described in the following sections.
Service providing organizations (SPO) comprise the next layer through which users are provided with connection services. SPOs will serve various numbers of individual users and organizations depending on their physical facilities.
The outer layer is composed of individual (single) users, user organizations, or organizations' networks that are to be connected to TR-NET backbone. These users located at the outer layer will be served by SPOs.
NCNC is planned to exist and be organized around following list of major policies and responsibilities:
The overall funding model is planned to be based on successive implementation of a two-part funding model. The first part of the model aims at performing a transition from governmental subvention to self-funding. The second part of the model is purely a self-funding model that is planned to be the ultimate model for funding the country-wide network in Turkey.
Initial investment budget will be required for the first year to construct the country-wide backbone at moderate speeds and to upgrade international bandwidth. This budget is estimated to be nearly one million US dollars for a 256Kbps backbone and 256 Kbps international bandwidth. Table-1 shows major cost items for the initial investment budget. As it is seen from Table-1, this budget includes operational cost items, such as line costs for the backbone and the international connection. These cost items are to be incurring in the following years. However, during the following years - also known as the transition period -, parts of these costs will be collected from end-users through service providers (SPOs). Hence, the total amount of subvention from the government is expected to decrease during the transition period.
Table-1. NCNC Initial Investment Budget
Communication 335,000 Equipment Other Hardware and 50,000 Software Line Costs for 140,000 Backbone International Line 320,000 Upgrade Cost Maintenance of H/W 40,000 and S/W Personnel 110,000 Other 25,000 T O T A L 1,020,000
Table-2. NCNC Operational Budget
Line Costs for 140,000 Backbone International Line 320,000 Upgrade Maintenance of H/W 40,000 and S/W Personnel 110,000 Research and 100,000 Development Other 25,000 T O T A L 735,000
Today, there are nearly 3,000 registered hosts in 100 zones, and 277 (allocated as only C-Class) registered IP networks in '.tr' domain. It is roughly estimated that there are around 10,000 to 15,000 daily users accessing TR-NET facilities from within the country. Using these figures, one can roughly estimate that the average cost per user would be some amount between 65 to 100 USD/year, while average cost per IP domain would be between 2,500 and 3,000 USD/year. This would of course necessitate a balanced (nearly uniform) bandwidth allocation among users of different IP domains. As it has been discussed earlier sections, these estimated basic costs are to be reflected to the user community as value-added services through SPOs.
NCNC is currently trying to complete the pricing structure that is to be applicable to SPOs. Pricing structure is based on volume charging. Depending on the type of connection, users will be charged either in accordance with (i) number of bytes flowing through (e.g., non-dialup connection), or (ii) number of minutes lapsed since the beginning of the connection (e.g., dialup connection).
The fundamental idea here is to have a profit margin only for improving the core services for SPOs (for better backbone bandwidth and topology, improved international connectivity and bandwidth, better technical support, etc.). Once SPOs begin serving the user community, funds will be flowing from users to SPOs, then SPOs to NCNC. As these funds become significant, governmental support (subvention) will garadually be decreased in proportion with the collected amount of funds. A rough estimation (an educated guess) suggests that the model will become a self-funding one within three years after deployment.
NCNC is planning to use a research and development budget item for funding development efforts on related software and hardware. It is our belief that development of some hardware and/or software will be necessary for at least providing user community with a set of homogeneous services through supporting SPOs.
In addition to basic connectivity services, some SPOs will provide a set of advanced value-added services, such as, construction of Web servers, allocating computing resources, running name servers, mail servers, specialized bulletin board services, etc. NCNC is not in a position to estimate basic unit prices for such services. However, unit prices for basic connectivity to TR-NET backbone are to be set by NCNC.
SPOs will charge their users for the basic connectivity services by putting a profit margin on top of the basic unit prices set by NCNC. The basic pricing structure that NCNC will use for charging SPOs is a simple formula:
cost = b + ai ( xi - xi-1) + aj (xj - xi); where
b > 0; ai >= 0; aj > ai > ai-1 >... > a1.
xi = 0, m, n, ... , k ; m > n > ... > k.
xj > xi.
where xi is measured in terms of (i) number of bytes, or (ii) number of connection minutes; b is the fixed cost, and ai is the slope of the pricing curve. The pricing curve is shown in Figure-7.

Figure 7. Pricing Curve
In fact, cost is constructed as a discrete function, because there will be a few ranges of traffic volume to which different rates are applicable, that is the value of a changes for different ranges of xi values (e.g., a1 for x0 x x1, a2 for x1 x x2, and a3 for x2 x; and a1 a2 a3 ...). In addition to this, total cost will have an upper limit for certain types of connections.
The fixed cost, b, for any connection from a SPO to TR-NET backbone is determined in proportion with the cost of initial efforts spent by NCNC, and the fixed equipment cost (e.g., additional router port). Another component of b is the license fee to be paid to NCNC by the SPO. For an SPO to determine its basic unit charges, other organizational expenses could be considered as part of this fixed cost.
The unit prices as well as quality and quantity of services offered by SPOs will determine their success in the marketplace. However, some of these quality and quantity parameters will be observed by the NCNC and verified against contractual commitments of SPOs, for the sake of user satisfaction. Some parameters are:
Charging the users by SPOs may depend on the same charging model that NCNC uses. That is, the user pays a fixed amount for getting a connection, and additionally, pays another amount proportional to his/her usage. SPOs may optionally use other charging mechanisms as part of their promotion strategy.
Today, it seems that international access is more important for the user community than intra-country communication. However, as network-based national information services become more prominent, current users will be more dependent and concious on TR-NET backbone, as well as new users joining the network.
An important asset for realizing the establishment and operation of a large scale backbone network is the accumulated technical know-how. This know-how accumulation, especially in last few years, is considered to be significant.
Today, in addition to universities and research institutions, there are more than 100 institutions and more than 1000 individual users accessing TR-NET facilities. Total number of users are estimated to be around 10,000. Current demand trend shows that the number of institutional and individual users will exponentially increase in the very near future.
Based on its significant experience in last three years, TR-NET organization is currently offering a model of organization and funding for network services in Turkey. The proposed model of organization places National Center for Computer Networks (NCNC) at the core of the model. NCNC is expected to operate TR-NET backbone and sell connection services to Service Providing Organizations (SPOs). SPOs will in turn reach to the user community and sell value-added services to them. The model does not propose a fully privatized structure, because central planning is considered to be very important for healthy development of computer networks in the country.
As a result of its national networking policy, TR-NET organization, and its candidate successor, NCNC aim at diffusing networking and information technologies among all sectors of the country. NCNC, while trying to achieve its goals, will consider those lessons that were learnt over the last ten years.
[1] Genç F.P, Özgit A, Önder S., 'Use of European Academic and Research Network (EARN) in Turkey', 5. Bilgisayar Kongresi, Istanbul, 1988.
[2] Özgit A., Yilmaz T., TÜBITAK EEE-AG DPT/64 Project Proposal, METU, Ankara, 1992.
[3] Özgit A., 'Internet: Availability and Usage in Turkey', Bilisim'93, Istanbul, 1993.
[4] Özgit A., and et.al., 'A Proposal for a National Policy on Computer Networks in Turkey', Biliþim'94, Istanbul, 1994.
[5] Özgit A., and et.al., 'Turkey's Networking Infrastructure: Past, Present, and the Future', Proceedings of JENC6, Tel-Aviv, Israel, May 1995.
Kürsat Çagiltay is a graduate of Mathematics Department, Middle East Technical University. He holds an M.Sc. degree from Department of Computer Engineering of Middle East Technical University. Mr. Çaðýltay is the manager of Networking in METUCC (METU Computer Center). He is a member of the executive committee of TR-NET.
Erdal Taner is a graduate of Computer Engineering Department, Middle East Technical University. Mr. Taner is the manager of Information Services in METUCC (METU Computer Center). He is a member of the executive committee of TR-NET. Mr. Taner's professional interests are: computer networks and communication technologies.