SI-Based Scheduling of Parameter Sweep Experiments on Federated Clouds - High Performance Computing - page 28

SI-Based Scheduling of Parameter Sweep Experiments
on Federated Clouds

Elina Pacini 1 , 3 , Cristian Mateos 2 , 3 , and Carlos García Garino 1

1

ITIC - UNCuyo University, Mendoza, Mendoza, Argentina
{epacini,cgarcia}@itu.uncu.edu.ar

2

ISISTAN - UNICEN University, Tandil, Buenos Aires, Argentina
[e-mail]

3

Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)

Abstract. Scientists and engineers often require huge amounts of computing
power to execute their experiments. This work focuses on the federated Cloud
model, where custom virtual machines (VM) are launched in appropriate hosts
belonging to di ff erent providers to execute scientific experiments and minimize
response time. Here, scheduling is performed at three levels. First, at the
broker level , datacenters are selected by their network latencies via three poli-
cies -Lowest-Latency-Time-First, First-Latency-Time-First, and Latency-Time-
In-Round-. Second, at the infrastructure level , two Cloud VM schedulers based
on Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) for
mapping VMs to appropriate datacenter hosts are implemented. Finally, at the VM
level , jobs are assigned for execution into the preallocated VMs. Simulated exper-
iments show that the combination of policies at the broker level with ACO and
PSO succeed in reducing the response time compared to using the broker level
policies combined with Genetic Algorithms.

1 Introduction

Scientific computing is a field that applies Computer Science to solve typical scientific
problems. A representative example of scientific experiments is parameter sweep exper-
iments (PSEs) [13]. Running PSEs involves managing many independent jobs, since the
experiments are executed under multiple initial configurations a large number of times,
to locate a particular point in the parameter space that satisfies certain user criteria. In-
deed, users relying on PSEs need a computing environment that delivers large amounts
of computational power over a long period of time. A kind of parallel environment that
has gained momentum is represented by Clouds [14].
Executing PSEs on Clouds is not free from the well-known scheduling problem, i.e.,
it is necessary to develop e

cient scheduling strategies to appropriately allocate the
jobs and reduce the associated computation time. Moreover, in federated Clouds [3] it
is necessary to properly manage physical resources, when they are part of geographi-
cally distributed datacenters. Therefore, for the e

cient execution of jobs in federated
Clouds, scheduling should be performed at three levels. Firstly, at the broker level,
scheduling strategies are used for selecting datacenters taking into account issues such
as network interconnections or monetary cost of allocating VMs on hosts that compose