VI. Present Higher Speed Options
When the demands of individual users for sophisticated external resources involving graphics, sounds and video become so great as to overtax the telephone dialup links described in the previous section, one needs to consider ways to improve the bandwidth of that link. Even if all users at a given site are restricted to simple text-based applications, saturation of the low-speed link will occur as soon as enough people are using it to access resources external to the school. If individual users would be content with 2400 bits per second links, then a single link at 14,400 bits per second could handle only six simultaneous users. Modems often include compression software which enables them to double or triple this number, and the traffic of multiple users gets interleaved in an efficient manner, but there is still a finite, and not very large, limit to the number of users such a link can comfortably handle.
Since we are presuming a local area network at the school site connected by the link in question, provision of a higher-speed links accomplishes two things. First, as we have just been discussing, it allows for a higher aggregate bandwidth. That is, the total amount of data that users at the school can transfer over this link each second is increased. But the higher-speed link also allows each individual user to have access to a higher peak bandwidth. This means that for brief periods of time, in which others might not be making demands on the network link, an individual user will have access to the full bandwidth of the external link. This means that these individuals can be performing more complex tasks over the link at higher speeds than would be possible with a slower link.
The technology of providing high-speed data links is something of intense interest to the telephone and cable companies. Hence any of the points made in the present section and the one which follows it should be checked carefully against current practice in your own service area. We will try to list current common services in the present section. In Section VIII we adopt a more speculative tone and try to identify certain trends that we anticipate the industry will follow. Since this area is currently in considerable flux, we expect to revisit this topic in another year and update the content of these two sections at that time.
The architecture that we have in mind for the school site is the same as in the previous section. There should be a local area network populated by user devices which can receive and process IP packets. There should be local file servers, mail servers and information servers, which most likely will be running on a single piece of hardware. There should be a separate router which has a single connection to the Internet cloud. The speed of the local area network will typically be faster than that of the school's external link. Hence it's the external link that is likely to pose a bottleneck for data flowing to and from user devices in the school. The architecture that we are proposing allows for simple upgrades of external connectivity without any change in the structure of the school's internal network.
There are three elements to consider in pricing any of the connectivity options discussed below. These elements can be understood by referring back to the model of a single computer accessing a dialup service. The computer must have a serial port to which a modem is connected. The serial port is an example of the more general concept of a computer interface to which some external device or service can be connected. The modem is an example of the more general concept of some line adapter through which the computer interface connects to the service in question. The phrase "line adapter" is not part of the standard jargon, but we know of no other term which means "the gadget (like a modem) which goes between a data interface and the telephone company's equipment." Finally there is the phone line to which the modem is attached. In the more general case one will have line charges specific to the type of service being used.
Just as a serial port may support a range of possible speeds, other computer interfaces are able to support a range of speeds. If you have an older computer, you may find that its serial port does not support the speeds available with some of the newer high-speed modems, and you will have to purchase a new serial port in order to make full use of these capabilities. In the same manner you may have to purchase a faster interface on your router to be able to access some of the higher-speed services discussed below.
The simplest option for speeds higher than those available through SLIP links is to lease a private line from the telephone company. Physically these lines are made of the same material as your other telephone lines, but the wires are carefully tested and tuned, or "conditioned" in the jargon of the telephone company, to see that they can carry data at the rated bandwidth. This service is not cheap, because it may require the telephone company to string new wire to your premises if existing wires don't test up to specifications. Leased lines are available in a range of bandwidths, from 56,000 bits per second (bps) up to 1,544,000 bits per second. A 56,000 bps line typically costs around $200 per month, and a 1,544,000 bps line (known as a "T1" link) is about three times that much. Your local costs may differ from these figures, since the cost of any given installation usually depends upon the distance of your site from the telephone company's central office which provides services to your site. Roughly speaking, the cost of this type of technology is related to the product of the desired bandwidth and the length of the line that supplies this bandwidth. Compared to standard telephone service on the basis of cost per unit bandwidth, a T1 line may be a good bargain, but the total annual cost would be uncomfortable for most schools. A 56,000 bps leased line is cheaper, but not much of a bargain in relation to standard telephone links. This is because of the low cost of analog phone lines and the advances in modem technology, which are gradually erasing the speed differences between SLIP links and 56,000 bps leased lines.
The computer interface required for use with a leased line is known as a synchronous interface. Such interfaces are manufactured in a variety of speed ranges. Most units support 56,000 bps links, and there exist models which handle lower speeds or speeds up to 2 million bps. Typical costs of these interfaces are in the range of $500 - $1,000. The line adapter employed with lower-speed leased lines (i.e. 56,000 bps) is called a DSU. Higher speed links require a more complex device, known as a CSU/DSU. Typical costs for a DSU are around $500, while CSU/DSUs cost about $1500.
Under certain circumstances it is possible to eliminate the costs associated with "conditioned copper," which is the basis of the leased lines discussed above. When distances (as measured from each site to the telephone company's central office) are short, there is no need for conditioned copper, and any telephone wire, i.e., unconditioned copper, can be used to link these sites. Such lines are known as LADS lines and may be procured for private use at costs typically no greater than that of a standard telephone line. Using devices known as Short-Haul Modems in place of the standard DSU, one can pass 56,000 bps through these lines. The cost of a Short-Haul Modem is comparable to that of a DSU (about $500). Newer technology, which we will discuss in section VIII, allows one to obtain even higher data rates. The principal limitation of this approach has to do with distance, since sites connected through this service must each be within a mile or two of the same central office.
Fortunately, newer technology allows for services which are much more economical than 56,000 bps leased lines. Digital telephone service known as ISDN (the "D" is for "Digital") provides data rates of 128,000 bps at costs no more than double that of standard analog service. Unfortunately the telephone companies did not have connectivity to the Internet cloud in mind when they developed this service, and they typically price it with significant charges for connection time. Where such charges exist, one must often reluctantly dismiss ISDN as a viable option for network connectivity. Check with your local phone company to see what rates apply. Some enlightened telephone companies offer what they call "metropolitan area Centrex," which might allow your school district to avoid time charges for ISDN data calls within your district or region. Others are offering "flat rate ISDN," which does away with time charges for ISDN data service. In Section VIII we'll mention another way to avoid these charges by using a minor variant of ISDN connectivity.
To make use of an ISDN connection you need a synchronous interface on your router or routing computer. The price of synchronous interfaces (currently $500-$1000) is likely to fall as more computers are equipped with ISDN interfaces to operate voice-mail systems and other telephonic applications. The synchronous interface must connected to a line adapter known as a Terminal Adapter, which costs another $1500.
ISDN is not a very recent technology. Some countries have deployed it quite extensively, but its deployment in the American telephone system has been slow. As a result it is likely that newer technologies will supplant ISDN service, which was designed fundamentally for end-to-end connectivity, or what the telephone company would call a circuit. As we have seen amply in the previous sections of this paper, wide area Internet conductivity doesn't involve circuits; it involves packets passing through the Internet cloud. Hence one is interested in circuit-based services like ISDN only for access to a local Internet provider. From the provider's router onward, traffic flows as a stream of packets.
An example of the problems in scaling ISDN services can be seen by considering what it takes to gain higher bandwidth through this type of service. The ISDN Basic Rate Interface, which is what we have been discussing in this subsection, provides two data channels, each of which carries 64,000 bps, for a total of 128,000 bps. If one desires a higher bandwidth connection through ISDN, one simply adds more data channels, which means adding more ISDN lines to your site. While higher bandwidth is achievable through this means, there are no economies of scale to be had here.
The oldest of the packet-based services is something known as X.25 (pronounced "X dot 25"). Many school districts and state departments of education have X.25 networks in services, typically linking remote terminals to a distant large computer. The X.25 protocol was designed to overcome the problems associated with running data services over older analog phone lines, which required extensive error-checking at each stage along a given network link. In the modern context it is not a good match for IP-based services. Performance is significantly degraded when one tries to run an IP service over an X.25 network, and we do not recommend this approach although this expedient has been adopted at least temporarily in many areas.
The modern successor to the X.25 protocol is known as Frame Relay. The name refers to the concept of grouping together some number of bits of information and transporting the resultant "frame" as a unit across the network. Frame Relay supports a range of speeds from 56,000 bps to 1,5440,000 bps at prices competitive with leased lines. Since, as with other packet-based services, Frame Relay allows the phone company to have many people sharing the same infrastructure, this service is intrinsically cheaper to provide than are leased lines, and one can anticipate significant cost reductions if the service is widely adopted.
The line adapter used for Frame Relay is the same as for a leased line, i.e., a DSU for 56,000 bps or a CSU/DSU for higher speeds. The same types of synchronous interfaces also suffice, although your router must have the appropriate software support for this protocol. Fortunately, this protocol is currently supported by many brands of routers.
Switched Multi-megabit Data Service (SMDS) is another service that is available in many metropolitan areas. It is typically more costly than Frame Relay, since a specialized CSU/DSU is required as a line adapter for this service. The CSU/DSU for SMDS typically costs about $4500 (which is triple the cost of a CSU/DSU for Frame Relay or leased lines.) This service allows for data rates in the range from 1.5 million bps to 45 million bps.
The connectivity options discussed in the present section will typically be supplied by a local telephone company - or by a network service provider working in association with a local telephone company. There are a number of network service providers who offer Internet access for local area networks in businesses, government offices or educational institutions. Of particular interest to the schools are the regional networks, which were originally established to provide research centers with connectivity to the Internet. The regional networks often offer the lowest available prices for nonprofit institutions. There also exist a number of companies which do business on a regional or national level and which offer competing services. A school district developing Internet connectivity should shop among these various vendors for the best price and the level of support and service required by the district.
The prices quoted for line adapters in each of the options considered refer to the cost of a line adapter at a given school site. If the line in question runs from the school to the offices of the service provider, one must consider the cost of the line adapter at the other end of the line, which typically doubles the total cost to the school for this component.
It's hard to predict which of these service considered will be offered by service providers in your area. Leased lines are an old standby, so you can count on them, even if you may be put off by their cost. Frame Relay is being marketed effectively by many telephone companies, so you can also expect to see it supported by many network access providers. Unconditioned copper is probably too specialized an option to expect to find on the commercial market. But you may want to use it for those sites in your district where it can work.
You may want to combine several of the services described in the section to fit your school district's requirements in the most economical way possible. For example, you might connect adjacent sites with unconditioned copper, link these connected sites to a central administrative facility via Frame Relay and connect that central facility to your local network service provider through a leased T1 line. For larger school districts this sort of approach can result in considerable cost savings. Smaller districts may find it economical to adopt a similar architecture, using county agencies or Intermediate Units as their network hub.
There is obviously an embarrassment of riches when it comes to the issues of network design which have been discussed in the present section. Many of the relevant technologies are new and not widely deployed at present. This means that it's important to cross-check the recommendations you receive from any single vendor or service provider. Your district may want to employ consultants recommended by local universities or by your regional network provider. Through groups like the Consortium for School Networking you can contact other school districts with needs similar to your own and compare network designs. In the following section we provide a summary of the main points made so far in this paper. Then in section VIII we take a brief look at technologies which may not be available in your area at the present time but which may be slated for deployment in the near future. These are options to keep in mind before committing major resources to any of the services that we have discussed so far.