Timing planning can be divided into two categories: 1) what the real-time computing community thinks; and 2) what the planning theory community is thinking about. Category 1 is a subset of category 2. Most real-time practitioners are not aware of category 2.
The main difference is that category 1 assumes a relatively simple special case when the deadlines are either met or missed, and the missing deadlines are a failure, so the criterion for optimal planning meets all deadlines (the so-called βhardβ real-time). Earliest Deadline (EDF) is the most common deadline planning algorithm in Category 1. There is a tremendous amount of literature on scheduling dates for category 1 β for example, in the IEEE symposium in real time. A good book is Stankovic et al. Scheduling for real-time systems - EDF and related algorithms .
AFAIK, there are currently no COTS real-time operating system products that implement scheduling, in particular EDF. Several commercial products were taken (for example, DEC, IBM), but they refused because of various difficulties, such as integrating EDF with other resource management (for example, synchronizers, unplanned actions) in the OS while maintaining backward compatibility. The solution is scheduling (EDF and other algorithms) as an integral part of the OS from scratch. I know about the three real-time COTS OS products that did this, none of which entered the market for organizational reasons not related to the OS: DECs Libra, IBM OS / 2 for PowerPC (done in collaboration with DEC) and Open Software Foundation OSF-1 Mk7.3a (implemented in collaboration with DEC and IBM). Some research OSs developed and deployed with bare hardware (such as Jensen Alpha in the CMU) have been able to include scheduling. Alpha took full advantage of the freedom to enable arbitrary scheduling algorithms, including EDF and Utility Acrual. Other research OSes sought to expand Linux (see Jonathan Anderson's VA Tech ChronOS Project). ChronOS is limited to the Linux base, but also supports Utility Accrual accrual scheduling algorithms.
Category 2 covers the whole topic of timing planning in general, of which category 1 is an easier subset. In particular, category 2 recognizes the concepts of wounding and being late in relation to the deadline. Planning optimality criteria include minimizing the number of missed deadlines, minimizing the average delay, the minimum maximum delay, and many (any) others. Technically, category 2 minus the subcategory of category 1 is βsoftβ in real time, although real-time practitioners and even researchers use many different inaccurate and inaccurate descriptions of the term βsoftβ real-time. This is a schedule for category 1. However, it is more realistic and more widely applicable, used in many industries (for example, in the field of transport, manufacturing, etc.). There is an even greater volume of literature than for category 1. A good textbook is Pindo Plan tion: theory, algorithms, and systems .
source share