| dc.description.abstract | Since the emergence of the digital computer in the 1940s, computer architecture has been largely dictated by the requirements of mathematicians and scientists. Trends have thus been towards processing data as quickly and as accurately as possible. Even now, in the age of large scale integration culminating in the microprocessor, internal structures remain committed to these ideals. This is not surprising since the main users of computers are involved with data processing and scientific computing. The process control engineer, who turned to the digital computer to provide the support he required in his ever increasing strive towards automation, has had therefore to use these generalized computing structures. His basic requirements however, are somewhat different to those of the data processing manager or the scientific user. He has to contend with an inherent problem of synchronizing the computer to the real-world timing of his plants. He is far more interested in the response time of the computer to an external occurrence than he is to sheer 'number-crunching' power. Despite the trends in process control towards distributed computing, even the most advanced systems require a relatively large central processor. This processor is called upon to carry out a wide variety of different tasks most of which are 'requested' by external events. Multiprogramming facilities are therefore essential and are normally effected by means of a real-time operating system. One of the prime objectives of such a real time operating system is to permit the various programs to be run at the required time on some priority basis. In many cases these routines can be large - thus requiring access to backing storage. Traditionally the backing store, implemented by a moving-head disc for example is under the control of the real-time operating system. This can have serious consequences. If real-time requirements are to be met, transfer to and from the disc must be made as rapidly as possible. Also, in initiating and controlling such transfer, the computer is using time which otherwise could be avai1ab1e for useful, process-orientated work. With the rapid advancement of digital technology, the time is c1ear1y right to examine our present computer architecture. This dissertation explores the problem area previously discussed - the control over the bulk storage device in a real-time process-control computer system. It is proposed that a possible solution lies in the development of an intelligent backing-store controller. This essentially combines the conventional low-level backing store interface with a special purpose processor which handles all file routines. This dissertation demonstrates how such a structure can be implemented using current technology, and will evaluate its inherent advantages. | en_ZA |
