Abstract
Multiprocessor structures yielding the best performance in hard real-time applications when taking the objectives of feasible processor scheduling, inherent deadlock prevention, minimization of context-switching operations, and guaranteed reaction times as a basis are investigated. The considerations reveal that single processor systems in co-operation with devices specialized in carrying through operating system nuclei have advantages over classical von Neumann and symmetrical multiprocessor structures. Furthermore, it is shown that nearly optimal look-ahead algorithms for the virtual storage administation employing the code of entire tasks as paging elements are closely related to the deadline-driven scheduling.
Original language | English |
---|---|
Pages (from-to) | 761-769 |
Number of pages | 9 |
Journal | Software - Practice and Experience |
Volume | 16 |
Issue number | 8 |
State | Published - Aug 1986 |
ASJC Scopus subject areas
- Software