On the Feasibility of Decentralized Grid Scheduling

Abstract

Many authors recognize the limitations of hierarchical Grid scheduling in scalable environments, and proposed peer-to-peer solutions to this problem. However, most peerto-peer grid resource management systems allow only to discover available resources at the time of the request. We claim that peer-to-peer techniques have the potential for actual Grid scheduling, where each resource maintains a schedule of its future allocation to jobs. We present such a protocol, which additionally allows users to specify desired properties about the requested schedules.

Decentralized Scheduling Algorithm for DAG Based Tasks on P2P Grid

“…The prominent one in them is [19]. It first shortlists resources from the grid, using CYCLON [14] gossip protocol.…”

“…If there is no change in the value of dynamic entities even then after a fixed time interval information list will be refreshed. This causes extra network traffic [19]. However network traffic due to time-to-time updating of information list on neighbor nodes will be extremely low [19].…”

Enforcing Fairness in Asynchronous Collaborative Environments

“…A different kind of asynchronous collaborative environment is a peer-to-peer grid (Foster andKesselman, 1998, Weel, 2008). Here, collaboration happens when a node issues a computational job to be executed by some number of its peers.…”

“…First, when a job scheduling request is issued by one node requesting resources from its peers, the grid uses a decentralized scheduling algorithm to identify nodes that will be simultaneously available in the near future, and may be used to run this job (Fiscato et al, 2008). This algorithm initially selects a (random) set of nodes capable of executing the job together.…”

“…For example, peer-to-peer file sharing applications can improve everyone's download speed of a file under the condition that those users are willing to donate their resources to upload file contents (Cohen, 2003). Similarly, a service operator may use a collaborative content delivery network, which relies on the willingness of Web server administrators to help each other if one experiences a temporary overload (Pierre and van Steen, 2006); another possible method is to use grid computing, where system administrators are willing to contribute their resources in exchange for future use of global resources (Foster and Kesselman, 1998).…”