Citation Context

... replication and dynamic bundling [31]. In particular, replication – speculative execution of multiple copies of the same job – was shown to decrease BOT turnaround time in failure-prone environments =-=[12, 13, 23, 24, 35]-=-. Many of these works devise specific replication algorithms applicable in a certain setup. Our contribution is in the explicit separation of the mechanisms that implement these techniques from the po...

Citation Context

...tasks are assumed to be independent and atomic. In the literature, parallel applications whose tasks are independent are sometimes referred to as Bag-of-Tasks applications (BoT) (as in Anglano et al. =-=[4]-=-) or parameter-sweep applications (as in Casanova et al. [7]). Resource management systems for Desktop Grids mainly use pull-based scheduling (see Choi et al. [8, 9]). In pull-based scheduling, when a...

Citation Context

...f competing applications, and their failures due to crashes/reboots or unplanned departures from the Grid. In response to this need, various scheduling algorithms have been proposed in the literature =-=[2, 3, 11, 15, 16, 23]-=-, that typically attempt to minimize the makespan of BoT applications (that is, the time taken to execute all the tasks in a bag) in spite of resource heterogeneity, performance fluctuation, and failu...

Citation Context

...me definitively unavailable; the objective is to eventually process all work. In a similar RR n° 70298 A. Benoit, Y. Robert, A.L. Rosenberg, F. Vivien inria-00413977, version 1 - 7 Sep 2009 context, =-=[3]-=- aims at minimizing both the completion time of applications and the amount of resources used. There is a very large literature on scheduling divisible workloads on assemblages of computers that are n...

Citation Context

...ypes of architecture are also proposed, some of them inspired by peer-to-peer systems [8]. To cope with node volatility, several mechanisms have been proposed, such as checkpointing and job migration =-=[3, 5, 24]-=-. These mechanisms allow to efficiently manage computing resources that are likely to be reclaimed by their owners at any time. Similarly, Kondo et al. [13] emphasize the need for resource selection w...

Citation Context

...tasks are assumed to be independent and atomic. In the literature, parallel applications whose tasks are independent are sometimes referred to as Bag-of-Tasks applications (BoT) (as in Anglano et al. =-=[4]-=-) or parameter-sweep applications (as in Casanova et al. [7]). Such 3 applications are becoming predominant for clusters and grids (see Iosup et al. [18] and Li and Buyya [23]). While determining the ...

Citation Context

...ead of information collection, we assume that state information exchange is done by mutual information feedback [3]. Thus, when a machine j1 needs the local state information of another machine j2, then it sends a request message to j2 which in turn sends back a reply message. Both the request and reply messages may embed other local state information as dictated by the load balancing policy. The tasks are assumed to be independent and atomic. In the literature, parallel applications whose tasks are independent are sometimes referred to as Bag-of-Tasks applications (BoT) (as in Anglano et al. [7]) or parameter-sweep applications (as in Casanova et al. [8]). Such applications are becoming predominant for grids (see Iosup et al. [9] and Li and Buyya [10]). While determining the exact task execution time on a target machine remains a challenge, there exists several techniques that can be used to estimate an expected value for the task execution time (see Rao and Huh [11]). The policies considered in this paper exploit estimates on mean task execution times rather than exact execution times. Furthermore, in grids, tasks that belong to the same application are typically similar in their re...

Citation Context

...e master-worker (or “bag-oftasks”) scheduling problem in the face of such volatility in the context of desktop grid computing, either at an Internet-wide scale or with an Enterprise [16], [17], [18], =-=[19]-=-, [7], [20], [21], [22]. Most of these works propose simple greedy scheduling algorithms that rely on mechanisms to pick processors according to some criteria. These processor selection criteria inclu...

Citation Context

...tasks are assumed to be independent and atomic. In the literature, parallel applications whose tasks are independent are sometimes referred to as Bag-of-Tasks applications (BoT) (as in Anglano et al. =-=[4]-=-) or parameter-sweep applications (as in Casanova et al. [8]). Such applications have been observed to be the predominant applications in large-scale distributed computing systems, such as Desktop Gri...