Dynamic scheduling of lightpaths in lambda grids

A Framework to Achieve Guaranteed QoS for Applications and High System Performance in Multi-Institutional Grid Computing

“…It has been shown that scheduling jobs with given start times, execution times and deadlines is an NP-Hard problem even if we consider a non-shared resource model [12]. In previous papers, we have presented scalable algorithms for scheduling ARs for both nonshared [6,9] and shared [10] resource models.…”

“…The Scheduler component schedules jobs in such a way that QoS constraints of ARs are met while high utilizations of resources are maintained. Depending on the model of the resource used, the Scheduler employs either our algorithm for a non-shared resource model [6,9] or our algorithm for a shared resource model [10]. The function of Match Advisor is to asses the degree of fit of a particular job given the QoS requirements of the job and current state (schedule) of the resource and report it to MMC.…”

“…As by reserving resources in advance the user can be assured of an upper bound on the response time, ARs can be used for ensuring end-to-end quality of service. Just like other resources, Grid-enabled network resources can also be reserved in advance for a more predictable network delay [3,5,6].…”

“…The goal of the resource broker is to efficiently map jobs to different resources and the goal of the scheduler is to order jobs mapped to a particular resource to achieve specified time requirements. In earlier papers [6,9,10], we have shown how laxity in the reservation window can help improve resource utilizations of AR based scenarios by leaving the final scheduling decision with the scheduler. Laxity of an AR on a certain resource is the difference between its deadline and the time at which it would finish executing on that resource if it starts executing at its earliest start time.…”

“…Laxity of an AR on a certain resource is the difference between its deadline and the time at which it would finish executing on that resource if it starts executing at its earliest start time. Earlier, we have presented scalable algorithms for scheduling ARs mapped to a given resource [6,9,10]. In this paper, we focus on the mapping component as our results show that the algorithm for mapping ARs to resources, herein called matchmaking, can significantly affect performance.…”