RAMAMRITHAM, K., and STANKOVIC, J.A.: 'Dynamic task scheduling in hard real-time distributed systems', IEEE 127 Sofw., 1984, 1, (3), pp. 6.%75  Home/Search   Document Not in Database   Summary   Related Articles   Check

....monitoring run time constraints of tasks, because of the possibility of tasks finishing before their pessimistic worst case deadline. Using a multiprocessor system, dedicating one processor for scheduling and one or several processors for application task, was first adopted in the Spring project [20, 19]. The first use of delayed replication of data with guarantees of eventual consistency were reported in the Grapevine project [4] 5 CONCLUSIONS The DeeDS prototype represents several uniquely integrated advanced concepts from active databases, distributed systems, and real time database systems ....

K. Ramamritham and J. A. Stankovic. Dynamic task scheduling in hard real-time distributed systems. IEEE on Transactions, July 1984.

....or allocate unused computation time (slack stealing algorithms : 4] 8] 19] 28] when the task arrives, 9 without effecting the successfully scheduled tasks. In distributed systems, the idea is to use a uniprocessor scheme for local scheduling, and perform distributed scheduling [16] for tasks which are potentially subject to timing failures at run time. Such tasks are typically those that arrive at runtime or those that were statically scheduled to realize time deadlines, but had to be rejected due to the arrival and scheduling of a dynamic task with high priority. To ....

....were statically scheduled to realize time deadlines, but had to be rejected due to the arrival and scheduling of a dynamic task with high priority. To schedule a locally rejected task, the key idea is to search for a computational resource where it can be successfully allocated. Different schemes ([16], 18] 22] 23] 26] have been presented in the literature for this purpose. Further, worst case assumptions regarding execution times and communication delays have been made to provide hard guarantees. Path Based systems such as the AAW system (see sections 2 and 3.1 ) consist of multitudes of ....

K. Ramamritham and J. A. Stankovic, "Dynamic Task Scheduling in Hard Real-Time Distributed Systems," IEEE Software, July, 1984.

....chapters corresponding to the subsystem later in the document. Most theoretical multiprocessor scheduling problem formulations to find an optimal schedule, even for static scheduling, are NP complete[GJ75] So mostly heuristics are used. Most of the heuristics have a setup of spatial hierarchy [RS84], BS91] A global scheduler running on the master processor, assigns tasks to slave processors. The slave processors are running the local schedulers. Our approach of a peer based heterogeneous functional scheduling contrasts with that. Isolation of unpredictable external events from the ....

.... scheduling problem is two fold: 1) Map tasks to processors (processor allocation) 2) Perform task scheduling on each processor individually (time 34 allocation) There have been approaches where the multiprocessor scheduling is split between a global scheduler, and several local schedulers[BS91] [RS84]. The global scheduler would perform processor allocation, and possibly time allocation keeping in mind task dependencies across processor boundaries. The local schedulers perform time allocation being aware of tasks just assigned to one processor. In these approaches the scheduling is more ....

[Article contains additional citation context not shown here]

K. Ramamritham, J.A. Stankovic, "Dynamic Task Scheduling in hard real-time distributed systems", IEEE Software July 1984.

....i . The problem is to associate with nodes of the distributed system, a set of tasks (possibly empty) such that every task is guaranteed to be computed by its node completely before its deadline. Dynamic scheduling of tasks on distributed systems with hard real time constraints is discussed in [1, 2]. that use besides the local node scheduling, a distributed scheduling scheme based on a combined approach of focused addressing and bidding to schedule tasks to the nodes. In focussed addressing, if a task cannot be guaranteed at a node, it is sent to a selected remote node that is estimated to ....

....node. The surplus processing capacity of a particular node is suitably modified each time a task is assigned to it. As we shall see later, the list of processing capacity of nodes as maintained by the scheduler node helps it in assigning tasks to nodes. Thus, unlike the schemes described in [1] [2], a newly arriving task gets assigned to a node with not much communication overhead between nodes as the task is not resent to a third node. Further, the local schedulers of nodes are independent of each other and do not interact among themselves, as tasks are not preempted on one node and ....

[Article contains additional citation context not shown here]

K. Ramamritham, and J.A. Stankovic, Dynamic Task Scheduling in Hard Real-Time Distributed Systems, IEEE Software, July 1984.

.... Zho86] Alternatively, only significant load changes can be broadcast resulting in a decrease in network overhead [LiM82, Zho86] Other methods of exchange involve nearest neighbors [Wil83, KeL84] Stankovic has proposed a strategy that includes a bidding phase where negotiation takes place [RaS84, Sta85]. Some significant results of these previous studies are as follows. a) Load balancing is beneficial since load imbalance occurs frequently in a system with ten or more processors. b) Excessive state information to make a load balancing decision is not necessary and may be detrimental. c) Sink ....

K. Ramamritham and J. A. Stankovic, "Dynamic Task Scheduling in Hard RealTime Distributed Systems," IEEE Software, Vol. 1, July 1984, pp. 65-75.

....tasks, and not to develop an advanced heuristic for the scheduling of the tasks themselves. As proof of concept, we use a simple slack based dynamic scheduling algorithm to schedule the realtime tasks. This algorithm can be easily replaced by a more complicated algorithm for dynamic scheduling [25, 36, 34, 16]. The simplicity of our algorithm allows us to concentrate on studying and analyzing the incorporation of fault tolerance while scheduling real time tasks. Motivation Although many control applications are strictly periodic, there are several examples of systems in which aperiodic non preemptive ....

K. Ramamritham and J. A. Stankovik. Dynamic task scheduling in hard real-time distributed systems. IEEE Software, pages 65--75, July 1984.

....comparison, the focus of this paper is not to define new frameworks, but instead, to define models and methods to be used in such frameworks and to analyze their effect on the adaptive applications. Extensive research has addressed the problem of dynamic resource allocation for both the real time [1, 3, 4, 9, 15, 17, 31, 40] and the non real time [13, 23, 27, 34] domains, typically considering dynamic resource allocation in the context of load balancing. However, the methods developed in these studies do not fit our target application model. This is because our model assumes that the resource needs of a ....

....the application. This variability prevents us from using a periodic task model [15, 17] in which performance requirements are fixed throughout an application s execution, and therefore worst case needs have to be considered. It also prevents us from using a sporadic task model, as in the real time [9, 31, 40] or the non real time [13, 34, 23] domains, because of the high overhead of taking resource allocation actions at each task arrival. In addition, the specification of 1 Command, Control, Communications and Intelligence a real time parallel task, as needed for an application component, is either ....

[Article contains additional citation context not shown here]

K. Ramamritham and J. A. Stankovic. Dynamic Task Scheduling in Hard Real-Time Distributed Systems. IEEE Software, Vol. 1, No. 3, July 1984.

....In comparison, the focus of this paper is not to define new frameworks, but instead, to define models and methods to be used in such frameworks and to analyze their effect on adaptive applications. Extensive research has addressed the problem of dynamic resource allocation for both the real time [1, 3, 4, 9, 15, 17, 31, 39] and the non real time [13, 24, 28, 33] domains. The methods developed in these studies do not fit our target application model, because our model assumes that the resource needs of a time constrained task, even when generated by the same type of event, may vary throughout the execution of the ....

....the application. This variability prevents us from using a periodic task model [15, 17] in which performance requirements are fixed throughout an application s execution, and therefore worst case needs have to be considered. It also prevents us from using a sporadic task model, as in the real time [9, 31, 39] or the non real time [13, 24, 33] domains, because of the high overhead of taking resource allocation actions at each task arrival. Resource reallocation triggered by runtime variation of applicationneeds has received less attention. Previous schemes proposed for both real time [1, 4, 17, 32] and ....

[Article contains additional citation context not shown here]

K. Ramamritham and J. A. Stankovic. Dynamic Task Scheduling in Hard Real-Time Distributed Systems. IEEE Software, Vol. 1, No. 3, Jul., 1984.

....In comparison, the focus of this paper is not to define new frameworks, but instead, to define models and methods to be used in such frameworks and to analyze their effect on adaptive applications. Extensive research has addressed the problem of dynamic resource allocation for both the real time [1, 9, 10, 18, 22] and the non real time [7, 14, 17] domains. The methods developed in these studies do not fit our target application model, because our model assumes that the resource needs of a time constrained task, even when generated by the same type of event, may vary throughout the execution of the ....

....the application. This variability prevents us from using a periodic task model [9, 10] in which performance requirements are fixed throughout an application s execution, and therefore worst case needs have to be considered. It also prevents us from using a sporadic task model, as in the real time [18, 22] or the non real time [7, 14] domains, because of the high overhead of taking resource allocation actions at each task arrival. Resource reallocation triggered by runtime variation of application needs has received less attention. Previous schemes proposed for both real time [10, 20] and ....

[Article contains additional citation context not shown here]

K. Ramamritham and J. A. Stankovic. Dynamic Task Scheduling in Hard Real-Time Distributed Systems. IEEE Software, Vol. 1, No. 3, Jul., 1984.

....of existing research on resource allocation and reallocation is focused on algorithms that determine how to most effectively allocate or reallocate resources. There is an extensive literature on dynamic resource allocation, typically in the context of load balancing algorithms (for example see [8, 15, 21, 12, 18, 9]) Strategies typically focus on where tasks must be scheduled as function of available resources. More recent research has studied dynamic processor scheduling algorithms in multiprocessor systems[14, 13] and even algorithms for dynamic control of communication resources[16] in ....

K. Ramamritham and J.A. Stankovic, "Dynamic task scheduling in hard real-time distributed systems'," IEEE Software, Vol. 1, No. 3, pp. 65-75, July1984.

....equivalent to the state before its execution. The dynamic concurrency control and task scheduling policies [28, 29] offered by CHAOS arc jointly address the potentially unpredictable operating conditions of complex real time applications. These policies must themselves be predictable [24], which implies that arbitrary delays due to locking [12] in concurrency control protocols or due to cascaded aborts in timestamping based protocols are not acceptable. The primary motivation for the development of CHAOS arc is our experience with real time applications that were or are being ....

K. Ramamritham and J.A. Stankovic. Dynamic task scheduling in hard realtime distributed systems. IEEE Software, 1(3):65--75, July 1984.

....reclaims resources to improve performance, and dynamically attempts re guarantees when resources are reclaimed. A performance study is accomplished via simulation. Our new algorithm is compared against the standard earliest deadline algorithm and a guarantee based earliest deadline algorithm [Ram 84] We compare the algorithms under widely varying conditions with respect to load, arrival rates, value distributions, allowed tolerances, and actual versus worst case execution times. The new algorithm significantly outperforms these baselines in all tested situations. The paper is organized as ....

....deadline scheduling has potentially catastrophically poor performance in overload. Various research has been performed attempting to rectify this problem. We briefly present EDF work related to overload as well as some general work on overload in real time systems. Ramamritham and Stankovic [Ram 84] used EDF to dynamically guarantee incoming work and if a newly arriving task could not be guaranteed the task was either dropped or distributed scheduling was attempted. All tasks had the same value. The dynamic guarantee performed in this paper had the effect of avoiding the catastrophic ....

K. Ramamritham and J. Stankovic, "Dynamic Task Scheduling in Hard Real-Time Distributed Systems," IEEE Software, pp. 65-75, July 1984.

No context found.

RAMAMRITHAM, K., and STANKOVIC, J.A.: 'Dynamic task scheduling in hard real-time distributed systems', IEEE 127 Sofw., 1984, 1, (3), pp. 6.%75

No context found.

Ramamritham K., and J. Stankovic. "Dynamic Task Scheduling in Hard Real-Time Distributed Systems." IEEE Software. IEEE Society, July 1984, pp. 65-75.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC