Architecture and Costs of Connecting Schools to the NII
MR. RUSSELL I. ROTHSTEIN, Research Assistant and DR. LEE McKNIGHT, Principal Research Associate MIT Research Program on Communications Policy Massachusetts Institute of Technology Cambridge, Mass. Universal connection to the National Information Infrastructure (NII), it is presumed, will facilitate educational reform within schools and communities. Several developments suggest this: The federal government is committed to having every classroom in the U.S.A. connected to the NII by the year 2000; A number of telephone and cable companies have announced plans to connect schools in their service areas at low or no cost; Modern, high-speed networks have been installed at a number of progressive, pioneering K-12 schools; The Internet, a global network of networks that connects to an abundance of educational resources, is experiencing phenomenal growth. However, to date, there is relatively little known about the costs for connecting schools to the information infrastructure. Even though exact costs are unknown, they are expected to be significant. To reduce these costs, a variety of approaches are being tried. Some states have implemented the cost-saving strategy of purchasing telecommunications equipment and services for all schools in the state. For example, North Carolina has saved its schools 20% - 50% on certain items. Some states have passed legislation permitting the state Public Utility Commission to set preferential or fixed intra-state rates for educational institutions. Our research suggests that a number of programs would have a significant impact on the total costs of connecting to the NII. For instance, if all schools coordinate purchasing at the state level, cost savings will exceed $2 billion. Colleges and universities often have the resources to provide technical support to K-12 schools. If a nationwide program were instituted, potential savings would be $800 - $1,800 million. If schools were given free Internet connectivity, the total annual costs for school Internet connections would be reduced from $150 - $630 million. But costs for telecommunications lines and services represent only 11% of the total costs, and hence reductions will have a limited impact. Finally, as the costs of networking schools is better understood, a new question arises: How will these costs be financed? Many states have programs to fund networking in schools. The federal government has a role, although it must become more flexible and coordinated. However, as Vice President Gore continues to state, the NII will be built by the private sector. A number of states have initiated cooperative ventures between businesses and schools. An expansion of these programs may be the key for successfully connecting K-12 schools to the NII. But connection alone is not enough: we report below on our finding that support and training together comprise 46% of the total costs of networking schools. Scope of K-12 Networking The model of school networks presented follows the Internet-networking model, by which schools have digital, data connections that transmit and receive bits of information. The models do not include analog video point-to-point networks or voice networks and voice-mail systems. Audio and video functions are possible in digital format over the Internet, but many schools will still use separate video and voice networks. Costs for these systems are important, but are not covered in this article. It should also be noted that although voice and video networks have been separated out from this report, schools should not consider these three types of networks to be wholly distinct. Some schools have integrated their voice and video networks with their data network. Sharing resources between the multiple networks can be effective in providing significant cost savings. At a basic level, it must be understood that as a school installs a LAN and puts computer data connections in every classroom, there are little added costs to also concurrently install other types of connections, including telephone lines. Technology Standard for Connecting to the NII As described in Information Infrastructure Task Force (1994), the NII "promises every ... school ... in the nation access anywhere to voice, data, full-motion video, and multimedia applications... Through the NII, students of all ages will use multimedia electronic libraries and museums containing text, images, video, music, simulations, and instructional software." The following requirements outline what the model presumes is needed in order to have full connection to the NII: A LAN within the school with connections to multiple machines in every classroom. The power of the network is greatly enhanced with more access points throughout a school. A classroom with one connection is not conducive for use of network applications in a class of 20 or 30 students. Telecommunications will not be a tool for systemic educational reform until network connections are replete throughout a school. A connection from each school to a community hub. From two to ten schools should connect to a single hub, depending on the size of the schools. In most cases, the hub will reside at the district office. However, where there are many schools in a single district, then schools should be clustered into sets of four to six. Each of these school-clusters will have a group hub, probably at the district office, which will contain the center of the network for those schools. The rationale for the use of this architecture is described below. A connection between the school LAN and the district office hub. With this configuration, every classroom has a connection not only to every other class in the school but also to the central district office. A connection from the school district office to a community-, state-, or nationwide Wide Area Network (WAN). This link will allow all schools to connect to the WAN. The Internet is a good example of such a WAN and will be used throughout this report as a model and precursor for the NII. Sufficient bandwidth for these connections. With a high-bandwidth connection, users in schools can make use of graphical applications (like Mosaic) and limited video service (like CU-SeeMe and MBONE). For most districts, the minimum bandwidth, or bit-speed, that will support these services is 56,000 bits per second (56 Kbps). Thus the connection between the school and the hub must be at or above this level. For the connection from the district office to the Internet, a broader pipeline is necessary because all of the schools in the group connect to the Internet through this line. The suggested minimum bandwidth for this connection is 1,500,000 bits per second (1.5 Mbps), also known as a "T1 line." Symmetric, bi-directional access to the WAN/Internet. It is important that the connection to the school allows information to flow both in and out at the same rates. In this way, students can become both consumers and providers of information over the network. Adequate remote dialup facilities. With a sufficient number of modems and phone lines, faculty, students and parents can gain access to the school system remotely on weekends and after school hours. Use of established and tested technologies. Schools have benefited most from mature technologies that have been well tested in the marketplace. Cutting-edge technologies have not been as successful in schools due to their inherent instability and the large amount of resources required to support them. The models assume the use of mature technology and transmission media. Newer technologies such as wireless and coax-fiber hybrid systems are not considered in this study. However, given the rapidity of technological change and marketplace evolution for networking products and services, wireless and cable alternatives should be evaluated in future research. Architecture of the District Network The basic network architecture for these models follows the "star" configuration, similar to that used in the nationwide telephone network. In the phone network, residential telephone lines in an area are directly connected to a single district office. In the school network, each school building is connected to the school central hub. In most cases, the district office will serve as the central hub. However, in cases where there are very few or many schools in one district, then alternative sites must be chosen. The rationale for adopting this architecture is that when many schools are connected through a single hub, then costs can be aggregated among them. This gives schools stronger purchasing power as equipment purchases are aggregated by the school district for volume discounts. It also allows schools to share resources-such as the data line to the Internet, training programs and full-time support staff-that each school might not be able to afford individually. Thus, there are costs both at the school and at the district level for networking schools across the country. A second rationale for adopting a star architecture is that it is confluent with the administrative/bureaucratic design of the school system. Individual schools report to a school district office, which in turn reports to state education offices. In the network design, schools connect to the district office hub, which in turn connects to a statewide (or national or global) network. Cost Areas The cost models presented in this article include four types of costs: hardware, training, support and retrofitting. Items included in these categories are summarized: Hardware - Wiring, routers, servers and PCs, including installation, maintenance and service of the hardware and telecommunications lines. Training - Training of teachers and other school staff to use the network. Support - Technical support of the network. Retrofitting - Modifications to the school facility to accommodate the telecommunications infrastructure. This may include costs for asbestos removal (where applicable), electrical systems, climate control systems, added security (locks, alarms, etc.) and renovation of buildings to accommodate network installation and operation. A cost area not included in the models is educational software. "Freeware" versions of many popular Internet applications exist, however, other educational software may be desired by particular schools. Economic analysis of such software costs and their evolution in network scenarios is needed.
Four Different Technology Models Following is a brief description of four ways to achieve connectivity to the NII. Inherent costs are noted; details are provided later in this article. Model One: Single PC Dialup This model represents the most basic connectivity option for a school. The school has no internal LAN within the building. There is a single connection to the district office over a modem and standard phone line. Only one person may use the connection at any time. Model Two: LAN with Shared Modem The difference between this model and the former one is the existence of the LAN within the school. By connecting the modem to the LAN, every computer on the network has access to the Internet. However, this model supports only a few users at a time, since it is limited by the number of phone lines going out of the school. As in the first model, users of the system can only utilize text-based applications over the Internet (e.g. e-mail, telnet, gopher). In this model, there is now a cost for the LAN. This model assumes the use of copper wire (category 5) as the medium for the network since it is the most affordable and scaleable option for schools in 1994. The costs for the wiring and network cards run $100 - $150 per PC connected. Including costs for the accompanying hardware and labor, costs per PC are $400 - $500. Therefore, for the school model with 60 - 100 connected PCs (3-5 PCs per classroom @ 20 classrooms), the total LAN costs are $20,000 - $55,000. Model Three: LAN with Router The main difference between this model and the former one is the existence of a router in place of the modem. With the router, multiple users of the LAN may access the Internet concurrently. Again, people can use text-based applications over the Internet, but have no real-time access to video or graphics. Since the router allows multiple users of the system, there is an opportunity to expand the entire network infrastructure. With this infrastructure, it is reasonable to support one PC in every classroom. Therefore, there is a requirement to purchase 15 additional PCs for the average school to use in addition to its small initial stock of TCP/IP-compatible machines. It is assumed that the purchasing of these PCs is done at the district level in order to negotiate better rates ($1,000 - $2,000 per PC). Support and training costs are higher since there are additional users of the system. There are additional dialup lines required to accommodate remote access. There are also significant retrofitting costs for the electrical system, climate control system and enhanced security. Model Four: The Preferred Strategy LAN with Local Server & Dedicated Line The primary difference between this model and the former one is the existence of a file server at the school. (See Figure 1.) The on-site server allows much of the information to reside locally at the school instead of at the district office. This feature provides better performance since more data d'es not need to be fetched over the network. Additionally, the local server allows school administrators to exercise greater control over the information flows in and out. Higher speed links from the school enable the use of limited video, graphical and text-based network applications. In this model, virtually the entire school is supported on the network. As a result, the training program is extensive and the support team is well staffed. Costs of the connection to the Internet are also higher due to the larger bandwidth. Significant retrofitting costs are incurred for the electrical system, climate control system and better security. A range of costs for an average size school to attain this level of technology is listed in Figure 2's table. Using the Models to Estimate Costs These four models are representations of the network technology used in schools. While a level of complexity and detail is omitted, the simplicity is helpful because they encompass broad cross-sections of network and school configurations. The models provide a clearer view of the costs and choices for networking K-12 schools. Using this model as a baseline for connecting to the NII, these figures then become indicative of the costs of connecting K-12 schools across the country to the NII. These numbers indicate that there will be $9.4 - $22 billion in one-time costs, with annual maintenance costs of $1.8 - $4.6 billion. At the per-pupil level, this is equivalent to $212 - $501 in one-time installation costs and an ongoing annual cost of $40 - $105. In this model, hardware is the most significant cost item for schools. However, most of the cost is for purchasing PCs. The value of PCs in schools g'es well beyond their use as networking devices. Therefore, the "real" costs for PC purchases should be allocated across other parts of the technology budget, and not only to the networking component. If this is done, then hardware costs for connecting schools to the NII drop considerably. If the high startup costs are amortized equally over a five-year period, then the breakdown of costs during the first five years, excluding PC purchases, is as shown in Figure 3's pie chart. Costs for support of the network represent about one-third. Support is a vital part of any successful implementation of a school network and its costs must be factored into the budget. Support and training together comprise 46% of the total costs of networking schools. Finally, it is important to note that the costs for telecommunications lines and services represent only 11% of total costs. This amount is lower than those assumed by much of the technology community, including telecommunications service and equipment providers. To put all of this into perspective: Total U.S. expenditures on K-12 education in 1992-93 totaled $280 billion. Total one-time costs for the fourth model described above represent 3% - 7% of total national educational expenditures. The ongoing annual costs represent between 0.6% - 1.6% of total national educational expenditures. Potential Impact of Initiatives on Costs Much more can be done by the government and the private sector to significantly mitigate the costs that schools face in order to connect to the NII. This section examines some possible programs and their impact on the costs to schools. Preferential telecommunications tariff rates are instituted for schools. Estimated savings: $89M - $218M (One-Time) (30% reduction) $39M - $150M (Annual) Estimated savings: $179M - $435M (One-Time) (60% reduction) $78M - $300M (Annual) All technology purchasing is done at the state level. Figures are based on an average of 30% discount across all 50 states. Estimated savings: $1.9B - $4.1B (One-Time) $45M - $189M (Annual) Universities or other institutions provide technical support to schools. It is assumed that schools will be able to function with 80% less support staff than would be required without university support. Estimated savings: $790M - $1.8B (One-Time) Teachers trained on their own time. If teachers agreed to attend classes on their own time, the only cost would be for the trainer. Estimated savings: $0 - $1.5B (One-Time) $0 - $300M (Annual) LAN installed by volunteers. If groups of parents and community members offer to provide labor at no cost, schools would reap significant savings. Estimated savings: $1.1B - $3.1B (One-Time) Personal computers are donated to schools. The success of a donation program depends on the quality of the equipment given. Schools will require fairly modern machines to run networking software. Donations of obsolete or incompatible equipment may be very costly to schools. Estimated savings: $5.1B - $10.2B (One-Time) Network routing equipment are donated to schools. This program is similar to a PC donation program. The savings are lower, however, since the routing equipment is less expensive. Estimated savings: $221M - $425M (One-Time) Network servers are donated to schools. Again, this is similar to the PC donation and router donation programs. Estimated savings: $370M - $1.5B (One-Time) Internet connectivity is free to schools. This could be arranged either by provision from an Internet service provider or from a local university or community college that has its own Internet connection. Estimated savings: $150M - $630M (Annual) Conclusions With a clearer picture of the costs for connecting schools to the NII, a number of conclusions may be drawn: The costs to network a school are complex. It is not simple to estimate the costs for a particular school. Costs for most schools will fall into a bounded range, but each particular school will vary greatly depending on its individual needs and characteristics. While this article seeks to put bounds on cost figures, the numbers are rough estimates at best. Network hardware cost is only a small fraction of the overall costs for connecting to the NII. Initial training and retrofitting are the largest one-time costs for starting the network. Costs for the wiring and equipment are typically not as high. Support of the network is the largest ongoing annual cost that schools must face. There are two major jumps in the costs to network a school. The first jump in cost arises when the school installs the LAN (see Figure 4). At that point the school and district must pay for installation ($20,000 - $55,000 per school) and employ full-time support staff ($60,000 - $150,000 per district). The second jump arises if and when a school decides to purchase computers for all students to use. The number of networkable PCs in 1994 is inadequate for most schools; hundreds of thousands of dollars would be needed to provide multiple PCs in every classroom. Also, many schools will need major electrical work (possibly $100,000+) to support the increased number of PCs. In the intermediate stages between these jumps, the costs are incremental and relatively small. Startup costs for the network increase at a faster rate than the annual ongoing costs as the network complex increases. In the less complex models, the one-time startup costs are 2-3 times the annual ongoing costs of the network. However, for the more complex models (models four and five,) the one-time costs are 5-15 times the costs to start the network. These differences are illustrated in Figure 5's graph. The divergence indicates that the most significant hurdle that a school will face is the initial investment costs in the network and computers. Dispensers of educational funding should be aware of this trend, so that they can help schools overcome this initial barrier. Schools should be given flexibility to amortize initial costs, in order to spread the burden over a number of years. Costs are significantly reduced when aggregated at the district and state levels. Schools stand to save a lot of money by pooling resources and purchasing power with other schools in the district and at the state level. When schools share a high-speed data link, or support staff, the per-school costs drop considerably. Schools in North Carolina and Kentucky, for instance, have saved 20% - 50% by purchasing services and equipment at the state level. Education initiatives of telephone and cable companies will have a small impact on the total costs to schools. The free telecommunications services currently offered by these companies are only a small piece of the total cost for successfully networking schools. However, schools would greatly benefit from the establishment of state trust funds or other cooperative efforts, as included in some plans. The benefits of these programs are more difficult to quantify, but are significant, nonetheless. They have the potential to garner support and consensus for connecting schools to the NII. In contrast, the highly touted free connections and free telephone services are not as valuable to schools. Further research on the costs of wireless and cable Internet access methods for schools is recommended to elucidate the costs and benefits of these approaches. In addition, the issue of software and equipment cost accounting require further analysis. We hope that this preliminary assessment of the costs of connecting schools to the NII can provide a point of departure for analysis of these and other more detailed models of NII connectivity for schools. Russell I. Rothstein is a Research Assistant and Lee McKnight is a Principal Research Associate with the MIT Research Program on Communications Policy, at the Center for Technology, Policy, and Industrial Development at MIT. E-mail: [email protected] [email protected]
This paper is based on a report prepared when Russell Rothstein was Visiting Researcher at the Office of Educational Technology in the U.S. Dept. of Education in 1994. Further work was supported by ARPA contract N00174-93-C-0036 and NSF grant NCR-9307548 (Networked Multimedia Information Services). The invaluable help of Dr. Linda Roberts, U.S. Dept. of Education, and Joseph Bailey, Thomas Lee and Sharon Gillett, MIT Research Program on Communications Policy, is acknowledged. References: 1. Rothstein, R. (1994), Connecting K-12 Schools to the National Information Infrastructure: Technology Models and Their Associated Costs, U.S. Department of Education Working Paper.
This article originally appeared in the 10/01/1995 issue of THE Journal.