gtpc3m0c | Concepts and Structures |
The user requests made of the TPF system are assumed to require very little computation, but considerable data manipulation. The online system is designed to prevent bottlenecks caused by queueing for system resources. The essence of the TPF system is to maximize performance for message driven applications on a given hardware configuration. Performance is viewed as a critical factor in the TPF system. This influences the techniques used to manage system resources. Generically, a control program or operating system defines application interfaces: control blocks, macros, supervisor calls, and so on. The TPF system has unique application interfaces that are influenced by techniques used to manage system resources.
The key elements of the TPF online system are:
The functions provided by these elements are outlined in the following sections.
The TPF main supervisor provides the resource management usually expected of a control program in direct control of physical facilities. The precise techniques used to accomplish this resource management are topics in TPF System Structural Characteristics. The principal functions performed by the main supervisor are:
Priority scheduling of input message processing is not done by the TPF system. An important processing concept that differentiates the TPF system from other operating systems is the way that it handles priorities and dispatches work (that is, an Entry).
Work already in progress, that is, active Entries, has the highest priority. Scheduling is accomplished using a limited number of work queues. Priority is determined by processing all of the Entries on a queue before processing the Entries on the next queue. For example, the queue of work that represents completed I/O operations that were requested by active Entries (called the ready list) has higher priority than, and therefore is processed before, the queue of work that represents new messages arriving at a CPC (called the input list).
In the TPF system, work scheduling is called dispatching.
In the TPF system, there are no access methods such as those found in MVS. Rather, channel programming is integrated into the system support of communication facilities, direct access storage devices (DASD), magnetic tape devices, and unit record devices.
Keypointing is the procedure by which system status is saved in the event that the system needs to be restarted because of a hardware or software malfunction. In other operating systems, this is usually called checkpointing.
The structure of the online data is identified during system generation. Data is dynamically generated; however, the facilities for storing the data are identified prior to the online operation.
There are two organizations of file structures within the online data:
A file request in the TPF system specifies a unique address whose exact form varies, but is always transformed into a record reference within an allocation scheme that spans the range of direct access storage devices (DASD), as shown by the horizontal allocation in Figure 4.
There are two categories of space within the online database:
An application program references a record in fixed file storage through the use of a record type (which is a value) and ordinal number (a relative number of a record within a specified record type). The TPF system translates this reference into a symbolic address.
A fixed file is analogous to a conventional multi volume data set organized for a direct access method. (Note that the term data set is an MVS term and is not used in conjunction with the TPF online database.)
This organization is used for several purposes:
Figure 4. File Allocation Strategies
In the TPF system, communications control provides the interface to data communication networks and message formatting services for applications. Because of the historical evolution of the TPF system, the control and support of communications is divided into two distinct areas:
On input, a message is in a format that is specific to the device from which the message was entered. Because input messages can be entered from various types of devices, the TPF system converts an input message with device-specific data to a common system format so that an application is able to process it regardless of its origin.
On output, the reverse is true; an output message must be converted from its common system format to be device-specific.
Additional detail about communications support is the subject of Data Communications.