Robert D. Carlitz

Department of Physics & Astronomy

University of Pittsburgh

Pittsburgh, PA 15260

Myron Lentz

Division of Computer Services

Pittsburgh Board of Public Education

341 S. Bellefield Avenue

Pittsburgh, PA 15213

Al MacIlroy

Techera

3907 Gresham Street #3

San Diego, CA 92109

(October 14, 1994)

SUMMARY

To assure the interoperability, reliability and maintainability of a school district's network, certain standards should be established and adhered to. The present article categorizes these standards and provides a list for designers and implementers to follow.

I. The Layered Approach

The goal of this article is to provide a simple and concise set of standards which school districts can use to facilitate the process of designing and implementing electronic data networks. Such networks are likely to be in increasing demand as new resources and educational applications are developed for the global Internet and as wide area networking increasingly permeates the society at large.

In planning the physical connectivity of a school district it is important to develop a broad view of the district's network architecture, including not only the infrastructure of the Local Area and Metropolitan Area network but also the set of applications that will initially operate over the network. This assures that the network will have adequate bandwidth for proposed applications, that these applications will be interoperable, and that sufficient funds will be available for all necessary hardware and software.

A great simplification in network design can be obtained if one thinks of the network as a set of levels, with each level isolated from the levels above and below it and communicating to these adjacent levels through a well-defined interface. This architecture allows one to design elements of the network without having to worry about unexpected interactions, incompatibilities or inefficiencies. The idea of a layered architecture has been carried to a formal extreme in the International Standards Organization's definition of seven layers of network structure. This definition is more formal than we will need in the present paper, and we simplify the ISO presentation by referring to three layers, each of which represents several of the formal ISO layers. The three layers that we will use are as follows:

* Physical Layer. This refers to the physical medium through which signals are carried, be it copper wire, fiber optic cable or wireless transmissions.

* Protocol Layer. This refers to the protocols used to encapsulate information on the network and present this information to applications running on devices attached to the network. These protocols define a set of rules which enable different entities on the network to communicate with each other.

* Application Layer. This refers to programs which run on computers attached to the network and provide specific tools or services to users of the network.

II. Physical Layer

Currently-used computer applications which operate on Local Area Networks in the school environment can be handled with inexpensive copper wiring. Present technology allows for operation at speeds of 10 million bits per second, and it is possible to install wiring which is capable of transmitting data at much higher speeds. A prudent recommendation is to use this type of wiring, known as Category 5 Twisted Pair wire. This choice combines economy of hardware and ease of installation with a reasonable allowance for future expansion.

The wiring plant should be a structured one, with a central wiring closet to which classroom or office runs return. Large sites will require multiple closets connected by backbones which can be constructed from either Category 5 copper or fiber optic cable. Sites of intermediate size may also be served economically with coaxial cable runs in some network segments.

Each classroom should have a minimum of three network drops. These drops can accommodate three devices, including classroom telephones as needed. Rooms which will require more devices can use fan-out hardware to accommodate as many devices as might be needed. The choice of three drops is a compromise between convenience and cost and is based upon experience in Pittsburgh and elsewhere with this wiring architecture. Some redundancy is desirable for ease of expansion, flexibility and added network integrity.

The preceding paragraphs refer to premise wiring or the structure of the Local Area Network within a given school. To connect schools together and provide access to central resources and the global Internet, one needs the fabric of a Metropolitan Area Network. Neither the precise needs in terms of bandwidth nor the precise solutions in terms of services can be specified at the present time with any great degree of confidence, but graphical applications require a minimal bandwidth on the order of 56 kilobits per second. Services which can provide this bandwidth include:

* Striped (parallel) modems over multiple voice-grade phone lines

* 56 kilobit leased lines

* Frame Relay

* ISDN

* Fiber optic lines

The layering concept allows one to mix these technologies at different sites in a network so as to obtain the required performance at the most economical cost. Unless one single technology can be obtained at a cost which is significantly lower than any competing technologies, one should plan to accommodate a mix of technologies at the physical layer of the Metropolitan Area Network, with the choice at each site matched to that site's needs and accessibility.

Graphical and video applications will eventually require higher bandwidth for many school sites. Here, too, there is a multiplicity of choices:

* Frame Relay at speeds of up to 1.5 megabits

* SMDS

* 1.5 megabit leased lines

* Cable TV

If one technology proves much cheaper than the others, it could provide a suitable choice for district-wide adoption. Otherwise one should anticipate a mix of technologies to form the fabric of the metropolitan area network.

III. Protocol Layer

This layer includes those protocols which define transport along the physical medium and protocols which present data to applications running on devices attached to the network. Transport over the Local Area Network described in the previous section is conveniently provided via the Ethernet protocol. Atop this protocol there sits a protocol which is independent of the physical medium. The choice of protocol at this level is simplified by the fact that the Internet is based upon a common public protocol known as TCP/IP. If students and teachers are to have access to resources on the Internet, then the Metropolitan Area Network and the Local Area Networks in the schools must support TCP/IP transport. This is the only commonly-used protocol which is suitable for application to wide area networks, and one can anticipate an evolution of popular proprietary protocols for local area networks to coincide with TCP/IP in the future.

Among the popular proprietary LAN protocols are IPX, which is used by Novell, and AppleTalk, which is used by Apple. While both protocols may be of use in specific LAN applications, they should not be extended over wide area connections, except through encapsulation in TCP/IP.

To isolate applications from the raw TCP/IP protocol, vendors have developed standards through which their applications receive data from the network. On the Macintosh platform the standard interface is provided through MacTCP, while on Windows machines the standard interface is known as Windows Sockets (WINSOCK). By enforcing these standards for each of these platforms, one can insure the interoperability of all network applications running on any given machine.

IV. Application Layer

Standards issues at this layer have to do with how applications handle data. We presume the existence of a networked environment, with all computers and peripheral devices connected to the network. Standard applications are available via file servers on the network, and maintenance of commonly-used software can be provided remotely. Local software can be added to individual classroom machines or installed on school-based servers. Virus-checking utilities protect individual machines from problem software, and the file structures for these machines can be rebuilt from the servers if their integrity is severely compromised.

Applications which support native mode standard file formats can exchange data and interoperate across the networked environment. This is an evolving area, as new applications are developed, but one can indicate a few general issues and a few specific issues which apply at present.

Among the general issues is that of how multi-media is handled. This question applies to mail, news and various network applications. The most widely-employed standard in this area is known as MIME (which stands for Multi-Media Internet Mail Extensions). It is reasonable to demand that mailers, news readers and network applications which support multi-media should all adhere to this standard. The standard is extensible, since it specifies external programs or "viewers" for any given multi-media file type. This enables it to accommodate a variety of text types, graphics formats, video and sound, as well as such specialized elements as databases and spreadsheets.

Specific issues relate to the user interface and file formats associated with specific user applications. A standard user interface can be specified in terms of a graphical display with a keyboard and a pointing device such as a mouse or trackball. Such interfaces are provided with all current commercial devices. Specialized interfaces should be provided for users with special needs such as voice synthesizers and mechanical assists.

Interoperability demands either that applications use a common file format or that they be provided with file conversion capabilities to and from commonly-used formats. The lowest common denominator for file exchange is that of ASCII text, and all programs should provide support for this type of format. One level up is a standard known as Rich Text Format, which allows for the specification of font information and attributes. This, too, should be supported wherever possible. User frustration can be reduced if a single product is deployed district-wide for each of the most common computing tools. These include

* Word Processing.

* Database.

* Spreadsheets.

Not all available software products meet all of the standards that we have specified at the present time. In order to provide students with access to products found in the typical commercial workplace it is therefore necessary to make some compromises in strict adherence to these standards. A convenient mechanism is to develop a district-wide list of current supported products. The list should indicate the extent to which listed products meet the standards stated here and the reasons for relaxing the standards in certain cases.

V. Attached Hardware

The previous sections already contain a number of implications for user devices to be attached to the school network. It should have a display capable of running a windowed environment, an Ethernet interface to the network and the processing power to run commonly-available applications. These requirements can be met with any of a number of models of Macintoshes or Windows PC's. Prudent recommendations for a minimal configuration of such machines are as follows:

* 8 megabytes internal memory

* 200 megabyte hard disk

* 800 600 pixel display with 256 colors

* 14" display screen

* Intel 486, Motorola 68040, or PowerPC processor

In addition to these user devices there are several other components required to make the network and its attached information resources function. These can be listed as follows:

* Servers. These machines provide file service, print service, information resources, and mail. A multi-tasking operating system is required to support this range of services. Typical devices for this task are RISC-based workstations running the Unix operating system. Other processor platforms may be adequate for smaller sites and other operating systems (notably Windows NT) may prove suitable for this purpose as their installed software base increases in size.

* Routers. These devices provide connectivity to the metropolitan area network. The choice of TCP/IP for MAN connectivity requires that the routers support this protocol. Multi-protocol support, including IPX and AppleTalk, can be a useful option but increases the required maintenance effort.

* Peripherals. This category includes printers, scanners, CD-ROM players, tape drives and other devices. Where standards exist in terms of file formats, these standards should be respected in most school district purchases. Relevant standards in this area include the PostScript page description language (for printers) and the Kodak Phot-CD format (for CD-ROM players). Networked devices are preferred for reasons of economy and flexibility. A high-volume, networked laser printer can, for example, serve a classroom more conveniently than multiple impact printers attached to individual computers. As with application software, this is an area in considerable flux, and standards should be reviewed on a regular basis.