A Hibernation Aware Dynamic Scheduler for Cloud Environments

Abstract

Nowadays, cloud platforms usually offer several types of Virtual Machines (VMs) which have different guarantees in terms of availability and volatility, provisioning the same resource through multiple pricing models. For instance, in the Amazon EC2 cloud, the user pays per hour for on-demand VMs while spot VMs are unused instances available for a lower price. Despite the monetary advantages, a spot VM can be terminated or hibernated by EC2 at any moment. In this work, we propose the Hibernation-Aware Dynamic Scheduler (HADS), to schedule applications composed of independent tasks (bag-of-tasks) with deadline constraints in both hibernation-prone spot VMs (for cost sake) and on-demand VMs. We also consider the problem of temporal failures, that occurs when a spot VM hibernates, and does not resume within a time that guarantees the application's deadline. Our dynamic scheduling approach aims at minimizing the monetary costs of bag-of-tasks applications execution, respecting its deadline even in the presence of hibernation. It is also able to avoid temporal failures, by using task migration and work-stealing techniques. Experimental results with real executions using Amazon EC2 VMs confirm the effectiveness of our scheduling when compared with on-demand VM only based approaches, in terms of monetary costs and execution times. It is also shown that our strategy can tolerate temporal failures.

A dynamic task scheduler tolerant to multiple hibernations in cloud environments

“…Likewise to HADS, several works in the related literature [25,21,24,19,29,27,28] propose, for monetary cost sake, the use, whenever possible, of spot VMs for scheduling tasks. On the other hand, as these works were conceived before December 2017, they cope with the termination/ revocation of spot VMs instead of their hibernation.…”

“…Such a heuristic was the first attempt to handle the hibernation of spot VMs, and results obtained by simulation showed that the hibernation problem is better handled with a dynamic approach. Thus, in [28], the authors present the first version of HADS. Contrarily to the current work, in that work, HADS tolerates only one spot VM hibernation, and, in case of migration, all tasks should restart from the beginning since no checkpointing technique is available.…”

“…To the best of our knowledge, the hibernation mechanism of the spot VMs is only discussed in our previous works [27,28] and in [10]. However, in [10], although the authors consider a scenario where hibernation-prone spot VMs can be used, they recognize that deadline constraints add complexity to the problem of resource provisioning, which should be evaluated in detail.…”

“…A preliminary version of HADS was presented in [28]. However, in that paper, we considered that a spot VM could hibernate just once and, upon migration, the task execution re-started from the beginning even if it had already executed part of its code before hibernation.…”

Avaliação dos Serviços de Armazenamento da Amazon Web Services para Gravação e Recuperação de Checkpoints

“…Diversas abordagens de tolerância a falhas foram propostas na literatura para lidar com as revogações do mercado spots [Teylo et al 2019, Yi et al 2010, Poola et al 2014. Uma das principais técnicas adotadaé o checkpoint and recovery (CR).…”