Authors:
Georgios L. Stavrinides
1
and
Eleni Karatza
2
Affiliations:
1
Department of Informatics, Aristotle University of Thessaloniki, Greece
;
2
Aristotle University of Thessaloniki, Greece
Keyword(s):
Scheduling, Complex Workload, SaaS Cloud, Performance, Modeling, Simulation.
Abstract:
As Software as a Service (SaaS) cloud computing gains momentum, the efficient scheduling of different
types of applications in such platforms is of great importance, in order to achieve good performance. In
SaaS clouds the workload is usually complex and comprises applications with various degrees of
parallelism and priority. Therefore, one of the major challenges is to cope with the case where high-priority
real-time single-task applications arrive and have to interrupt other non-real-time parallel applications in
order to meet their deadlines. In this case, it is required to effectively deal with the real-time applications, at
the smallest resulting degradation of parallel performance. In this paper, we investigate by simulation the
performance of strategies for the scheduling of complex workloads in a SaaS cloud. The examined
workload consists of non-real-time applications featuring fine-grained parallelism (gangs) and periodic
high-priority soft real-time single-task applications
that can tolerate deadline misses by bounded amounts.
We examine the impact of gang service time variability on the performance of the scheduling algorithms, by
considering service demands that follow a hyper-exponential distribution. The simulation results reveal that
the relative performance of the employed scheduling strategies depends on the type of the workload.
(More)