![]() (jointly referred to as rate-controlled EDF or RC-EDF) has been Schemes that have been proposed in the literature, earliest deadlineįirst (EDF) scheduling in conjunction with per-hop traffic shaping The provision of quality of service (QoS) in terms of end-to-endĭelay guarantees to real-time applications is an important issue inĮmerging broadband packet networks. ThisĬlass is introduced in this paper and may be of interest in other One of the proposed policies is based on aĬlass of non-preemptive policies that tracks preemptive policies. They also address the issue of designing policies that satisfy delayĬonstraints in a fair manner. Satisfiable by any policy, and they also have low buffer requirements. That are optimal in the sense of satisfying the constraints if they are For classes with delay constraints they provide policies Investigate the buffer requirements under three typical memoryĪllocation mechanisms, that represent trade-offs between efficiency andĬomplexity. Under the assumption of commonly accepted traffic constraints. Optimal in the sense of minimizing buffer and/or delay requirements Problem of characterizing and designing scheduling policies that are This paper is motivated by the need to support multiple serviceĬlasses in fast packet-switched networks. ![]() By developing the model as general framework for timing analysis we have also obtained some extensions of existing results about EDF and FPP. Not only FPP and EDF but also the classical first in first out or the last in first out policies fall in this category. We develop in particular a common approach for deriving response times bounds under policies that can be defined by priorities which are assigned to instances of tasks. ![]() We identify generic ideas and concepts such as majorizing work arrival functions, which allows to study scheduling policies separately from assumptions on tasks. In this report we propose a mathematical model based on trajectories to represent and analyze the scheduling of recurrent tasks on one processor. For the well studied earliest deadline first (EDF) and the fixed priority preemptive (FPP) policy, results are known under various assumptions about tasks. Timing analysis of real time scheduling policies is concerned with the analysis of response times, because real-time constraints impose that tasks must complete before their deadlines. Monotonicity properties are established for both policies Their deadlines before entering service and when they are not allowed toīe discarded. Policies exhibit exactly the same behavior for a fixed value of nīoth when customers are allowed to be discarded when they miss It is shown that these seemingly dissimilar Number in the system exceeds n, forces one customer from the ML arriving customers enter the ML queue and if the total Other policy, ML ( n )/ F, places the ML queue at customers finding n or more in the system enter theįirst-in-first-out (FIFO) queue which in turn feeds the ML queue. One policy,į / ML ( n ), places the ML queue at the front, Of unbounded size, managed in a first-in-first-out manner. N >0, managed using the minimum laxity policy, and another, Overhead by dividing the queue in two: one, of maximum size Minimum laxity (ML) scheduling policy without incurring its complete Both policies attempt to approximate the performance of the The behavior of two policies for scheduling customers withĭeadlines until the beginning of service onto multiprocessors is
0 Comments
Leave a Reply. |