No abstract
Abstract-Applications to be executed in Grid computing environments become more and more complex and usually consist of multiple interdependent tasks. The coordinated execution of such tightly or loosely coupled tasks often requires simultaneous access to different Grid resources. This leads to the problem of resource co-allocation. Efficient and robust scheduling algorithms have to be developed that can cope with the Grid's largescale distribution, a high number of competing and demanding applications, the inherent resource heterogeneity and the often limited view on resource availability. In this paper, we present two heuristic scheduling algorithms that are based on a well-known list scheduling algorithm and both support coallocation and advance resource reservation. Our first algorithm preserves the run-time efficiency of Greedy list schedulers while the second approach incorporates more sophisticated search techniques in order to achieve better results with respect to the performance metrics. Both algorithms have been implemented within a Grid simulation framework. An extensive simulation study was conducted to evaluate and compare the performance of both algorithms. It showed the general suitability of our enhanced list scheduling heuristics within heterogeneous Grid environments.
Abstract-The web is a crucial source of information nowadays. At the same time, web applications become more and more complex. Therefore, a spontaneous increase in the number of visitors, e.g., based on news reports or events, easily brings a web server in an overload situation. In contrast to the classical model of distributed denial of service (DDoS) attacks, such a so-called flash effect situation is not triggered by a bulk of bots just aiming at hurting the system but by humans with a high interest in the content of the web site itself. While the bots do not stop their attack until told so by their operator, the user try repeatedly to access the site without knowing that the repeated reloads effectively increase the web server's overload. Classical approaches try to distinguish between real user and harmful requests, which is not applicable in this scenario. Simply restricting the number of connections leads to very technical error messages displayed by the users' client software if at all. Therefore, we propose a mean to efficiently block connection attempts and to keep the user informed at the same time. A small subset of HTTP and TCP is statelessly implemented to display simple busy messages or relevant news updates to the end user with only few resources. In this paper we present the protocol subset used and discuss the compatibility problems on the protocol and client software level. Furthermore, we show the results of performance experiments using a prototype implementation.
Abstract-Resource reservations in advance are a mature concept for the allocation of various resources, particularly in grid environments. Common grid toolkits support advance reservations and assign jobs to resources at admission time. In such a distributed environment, it is necessary to develop carefully tailored failure recovery mechanisms that provide seamless transparent migration of jobs from one resource to another. As the migration of running jobs is difficult, an important issue in advance reservation, i.e., planning based, management infrastructures is to determine the duration of a failure in order to remap jobs that are already allocated to a currently failed resource but not yet active. As shown in previous work, underestimations of the failure duration and as a consequence the remapping of too few jobs results in an increased amount of job terminations. In order to overcome this drawback, in this paper we propose a load-based computation of the jobs to be remapped. A centralized and a distributed version of the strategy are presented, showing it is not necessary to have knowledge beyond the local allocation on the failed resource. The loadbased strategy achieves to effectively remap jobs while avoiding -inevitably inaccurate -estimations of the failure duration.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.