As highlighted by a number of recent articles in LSN, data communications are of increasing concern to libraries, particularly those wishing to use telecommunications to link remote terminals to a central computer and libraries seeking to interconnect remote systems for search and messaging access. A library administrator investigating data communications options will not venture far before encountering reference to the Open Systems Interconnect (OSI) Reference Model.
Until recently, data communication has lacked industry-wide standards, probably because computing has not been a regulated industry and communication among computer devices is a relatively recent development. All that purchasers have been able to expect is that all equipment supplied by a single vendor will be compatible with other equipment from that vendor. Because each computer vendor supports its own unique network structure, it has been impossible to interconnect systems supplied by different vendors. An overwhelming number of systems are able to communicate only with their genre. The development of standards to enable the linkage of equipment supplied by different vendors has appeal for both users and vendors. Users are not tied to a particular vendor, and because easier interconnection increases the use of distributed processing, vendors can access a larger market.
These perceived benefits fueled the development and approval, under the auspices of the International Standards Organization (ISO), of a network architecture model based on open systems interconnection as an international standard. The result if the OSI Reference Model (DIS 7496).
The first step in the development of OSI was to define all of the tasks required to permit two users to communicate across a network. The second step was to organize the tasks so that link functions were grouped together for most efficient interaction. The model is based on a layered architecture which segments the problem of systems linkage into functional groups or levels, each of which has unique attributes and --interacts with adjacent layers in the model. Each layer defines a discrete task: the layers build on each other to define operations of increasing complexity. The determination of the number of layers in the OSI model was influenced by a concern that they be sufficient to provide for the ready implementation of the structure, but not so many as to hinder development; and to minimize the number of interactions across layer boundaries. The resulting OSI model has seven layers, or levels and serves not only as a template for the design of network protocols, but has also become a powerful tool in analyzing and understanding network functions. The names and numbers of the levels of the model have become a part of the definition of communication products from long distances packet networks to local area networks.
OSI is not a specification of a particular network system design. Rather, the model is a framework, or skeleton, that permits standard interconnect procedures to be defined as protocols or rules which enable diverse users to interconnect and exchange information, regardless of their particular network or equipment vendor.
The model's only requirement is that- a system or device be connected to a physical communications medium through which access can be gained to one or more other systems. The medium can be as simple as a point-to-point communications line or as complex as interconnected private, local, and packet-switched networks. The device can be a computer, a terminal or anything else that has the intelligence necessary for information processing and communication. The interconnected systems may, but need not, be products of the same vendor.
In order of increasing complexity, the layers and their major functions are:
Physical layer
This is the lowest layer of the OSI architecture. It serves to permit the transfer of a bit stream over a physical circuit. The services defined by this level include acquiring, maintaining, and disconnecting the physical circuits that form the connecting path. Physical layer protocols handle the electrical and mechanical interfaces as well as the procedural requirements of the interconnection medium. These include connector type and voltage levels and the signaling procedures for control of the connection. The OSI model requires that the physical level be satisfactory for the transmission of "transparent data"; it cannot use procedures that restrict the characters that users can send across the connection. Typical physical layer protocols include the P8-2320, the P8-449 family, CCITT x.25 and X.21 facility interfaces, other CCITT V and X series recommendations, and media access protocols for local area networks.
Data link layer
Procedures in this level of the model are responsible for reliable delivery of link control data and user information over a point-to-point or multipoint link that has been established at the physical layer. Link layer protocols manage the establishment, control, and termination of error-free links; and control the flow of user data. Data link control protocols include the character-oriented binary synchronous communications (BSC) conventions, ANSI X3.28, DDCMP, and the more recent bit-oriented ADCCP and its international counterpart, HDLC.
Network layer
Procedures in the network layer select a route from among the available data links arranged as a network. This layer provides services for moving data through a network with concatenated data links and multiple routes available between points. These services include routing, switching, sequencing of data, flow control, and error recovery. Although some are duplicated at the link level, these functions are for network connections rather than data links.
Network layer protocols isolate the transport layer from concern for routing and switching. These protocols select and control logical paths and connections between networks user end points. Internet protocols, which control routing and recover between network nodes, and gateway protocols, which control data transfer between network, form the upper sublayers of this level of the model.
The CCITT X.25 packet layer is the most well-known network layer protocol.
Transport layer
This level of the model provides end-user-to-end-user assurance of information transfer; the user does not have to be concerned about the actual movement of the information. It is the highest layer directly associated with the movement of data through the network. This layer provides the higher layers, which represent the users of the communications service, with a universal transparent transfer mechanism. The transport layer is expected to optimize the use of available resources while meeting user requirements. It does this by segmenting or blocking messages that are not economical for network transmission.
Because transport protocols are responsible for ensuring the end-to-end integrity of the data exchange, they must bridge the gap between services provided by the underlying network and those required by the higher layers. Development work is still going on at this level. Simple transport layers will be used when a network provides a high-quality, reliable service; a complex transport protocol will be used when the underlying service is poor. In effect, the transport layer duplicates recovery mechanisms and other features that should have been provided by the lower layers.
Session layer
Services in this level coordinate the communications interchange between co-operating application processes. Sessions are established when one application process requests access to another. The session layer provides two categories of services: administrative and dialogue. An administrative service establishes and releases a connection between two presen-tation entities. After a session is established, dialogue services control and supervise the actual data exchange. Current session protocols include ECMA 75 and CCITT X.62, which is used in Teletex services.
Presentation layer
This level of the model serves to ensure compatible syntax among the communicating processes by adjusting data structures, formats, and codes. An application uses presentation layer services to properly interpret the information being transferred. It provides each user with an exchange dialog with the network (and therefore with other users); it is compatible with local operating and data format requirements. The services of this level include translation, formatting, etc. It also provides a means to request connections through the session layer and to terminate connections when communication is complete.
Work on formal presentation protocols has made significant progress; for example, recent work has aligned the many versions of videotex protocols into a single international presentation protocol.
Application layer
Services in this level of the model provide a window by which the user gains access to the communications services provided by the architecture. Since this is the highest layer of the OSI architecture, the application layer provides its services to the application process only. This layer is often identified as the "user," but it is more properly the user interface to the host computer's communication method. The application layer provides communication services directly to the user. These services include identifying the cooperating processes, authenticating the communicant, verifying authority, determining the availability of resources, and ensuring agreement on syntax. Applications do not reside in the layer; the layer is simply the means for the applications to gain access to the services provided by the communications architecture.
Work on the formal definition of application layer service descriptions and protocols will meet the needs of a specific applications such as libraries, banking, airlines, etc.
The OSI layers are also often described by their numerical position in the architecture. For example, the physical layer is Level 1, and the application layer is Level 7.
Previous communications systems have also been based on layered approaches. Computer manufacturers have offered layered network architectures to interconnect their own equipment lines. Two examples are IBM's System Network Architecture and Digital's DECnet. These architectures, however, are basically homogeneous because total compatibility occurs only within the computer product line of a single vendor. With the open systems architecture, the layered approached is used as a skeleton on which compatibility is built between heterogeneous or different computer systems.
