From: ahemavathy (ahemavathy_at_hotmail.com)
Date: Tue Jan 06 2004 - 22:02:11 PST
This paper describes "THE" - a multiprogramming system whose primary goal is to process smoothly a continous flow of user programs. The system has a hierarchial structure with 6 levels. Each level hides the implementation details of all the other levels lower than itself. The lowest level - level 0 - takes care of the processor allocation, processes interrupts and schedules or prioritizes interrupt handling. Above this level is level 1 which takes care of book keeping - the memory or disk access. Level 2 takes care of the allocation of the console keyboard i.e it makes it look like a private conversation between the user and a process. Level 3 takes care of the abstraction of peripherals. At levels 4 and 5 are the programs and the operators.
The system was designed with some objectives in mind such as 1) reduction of turn around time 2) economic use of peripheral devices etc. I was more interested in whether the objectives were achieved and if yes, some comparison of the multiprogramming system results versus a non-multiprogramming system results. Unfortunately the paper does not go into the observed benefits of the system over others.
The terms such as pages, segments (blocks), primary and secondary store are very much the same even today. He also has defined Random Access Memory although not explicitely. Also "semaphores" using which mutual synchronizations of parallel sequential processes are implemented were invented by the author.
The author's confidence about the system being flawless should be admired although we don't know how far true it is. Atleast from his perspective he has got all potential problems covered. He admits though that he did not pay attention to the pathology and easy debugging. They have followed a bottom-up testing procedure, i.e starting at level 0, the system was tested each time adding the next level only after the previous level had been throughly tested. He also emphasises on the importance of white-box testing.
This archive was generated by hypermail 2.1.6 : Tue Jan 06 2004 - 22:03:24 PST