A simple model to exploit reliable algorithms in cloud federations

Abstract

Exploiting resources belonging to multiple cloud providers in an efficient way is still an open issue for distributed computing. Scheduling algorithms based on heuristic, probabilistic, queue theory, or complex soft computing methods are suitable to tackle the heterogeneity and dynamism present in cloud federations. Nevertheless, the available brokering tools are focused on the deployment of services on-demand. The systems able to accomplish high throughput calculations, such as the pilot-job systems, do not support the inclusion of these algorithms due to their lack of adaptability. The recently implementation of cloud drivers for the GWpilot framework allows developers to profit from its flexibility, compatibility and scheduling features. Moreover, the framework allows the personalised characterisation of cloud resources that those algorithms require, overcoming their lack of trustworthiness in the information provided by the cloud services. In this work, a simple model together with a methodology to couple scheduling software with GWpilot is presented. To demonstrate the suitability of the approach, a legacy self-scheduler specialised on reliable executions in dynamic environments has been stacked and tested on the EGI FedCloud infrastructure with the Nagano legacy application.IaaS cloud providers are heterogeneous and their real availability dynamically changes through the time (Foster et al, 2008). Thus, the allocation of virtual machines (VMs) across multiple providers is a NP-complete problem (Garey and Johnson, 1979) that can be tackled with the sub-optimal approaches that soft computing or the operative research (Pinedo, 2005) can provide. For example, meta-heuristics (Gómez-Iglesias et al, 2010), evolutionary algorithms (Nesmachnow et al, 2010), or fuzzy logic (Saleh, 2013) have demonstrated their suitability for

Boosting Advanced Computational Applications and Resources in Latin America through Collaboration and Sharing

“…An exploratory work can be found in "A Simple Model to Exploit Reliable Algorithms in Cloud Federations." 13 All of these developments will converge into tools that combine dynamic checkpointing and fault-tolerance capabilities with cloud job resource managers over CPU and accelerators. Container management in virtualized HPC environments will also be studied.…”

AutoCompBD: Autonomic Computing and Big Data platforms

“…This paper studies the efficiency of data-intensive applications, and the penalties, in terms of power consumption, that are introduced by different degrees of heterogeneity of the virtual machines characteristics in a cluster. -A simple model to exploit reliable algorithms in cloud federations (Rubio-Montero et al 2016). The paper presets a simple model together with a methodology to couple scheduling software with GWpilot, a framework that allows the personalized characterization of cloud resources that those algorithms require, overcoming their lack of trustworthiness in the information provided by the cloud services.…”