Evaluation of gang scheduling performance and cost in a cloud computing system

The Journal of Supercomputing (Impact Factor: 0.92). 01/2010; 59(2):975-992. DOI: 10.1007/s11227-010-0481-4
Source: DBLP

ABSTRACT Cloud Computing refers to the notion of outsourcing on-site available services, computational facilities, or data storage
to an off-site, location-transparent centralized facility or “Cloud.” Gang Scheduling is an efficient job scheduling algorithm
for time sharing, already applied in parallel and distributed systems. This paper studies the performance of a distributed
Cloud Computing model, based on the Amazon Elastic Compute Cloud (EC2) architecture that implements a Gang Scheduling scheme.
Our model utilizes the concept of Virtual Machines (or VMs) which act as the computational units of the system. Initially,
the system includes no VMs, but depending on the computational needs of the jobs being serviced new VMs can be leased and
later released dynamically. A simulation of the aforementioned model is used to study, analyze, and evaluate both the performance
and the overall cost of two major gang scheduling algorithms. Results reveal that Gang Scheduling can be effectively applied
in a Cloud Computing environment both performance-wise and cost-wise.

KeywordsCloud computing–Gang scheduling–HPC–Virtual machines

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: A core idea of cloud computing is elasticity, i.e., enabling applications to adapt to varying load by dynamically acquiring and releasing cloud resources. One concrete realization is cloud bursting, which is the migration of applications or parts of applications running in a private cloud to a public cloud to cover load spikes. Actually building a cloud bursting enabled application is not trivial. In this paper, we introduce a reference model and middleware realization for Cloud bursting, thus enabling elastic applications to run across the boundaries of different Cloud infrastructures. In particular, we extend our previous work on application-level elasticity in single clouds to multiple clouds, and apply it to implement an hybrid cloud model that combines good utilization of a private cloud with the unlimited scalability of a public cloud. By means of an experimental evaluation we show the feasibility of the approach and the benefits of adopting Cloud bursting in hybrid cloud models.
    Proceedings of the 2013 IEEE/ACM 6th International Conference on Utility and Cloud Computing; 12/2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The cloud provider plays a major role especially providing resources such as computing power for the cloud subscriber to deploy their applications on multiple platforms anywhere; anytime. Hence the cloud users still having problem for resource management in receiving the guaranteed computing resources on time. This will impact the service time and the service level agreements for various users in multiple applications. Therefore there is a need for a new resolution to resolve this problem. This survey paper conducts a study in resource allocation and monitoring in the cloud computing environment. We describe cloud computing and its properties, research issues in resource management mainly in resource allocation and monitoring and finally solutions approach for resource allocation and monitoring. It is believed that this paper would benefit both cloud users and researchers for further knowledge on resource management in cloud computing.
    International Journal of Machine Learning and Computing (IJMLC). 02/2014; 4(1):31.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Multi-objective combinatorial optimization (MOCO) is an essential concern for the implementation of large-scale distributed modeling and simulation (MS) system. It is more complex than general computing systems, with higher dynamics and stricter demands on real-time performance. The quality and speed of the optimal decision directly decides the efficiency of the simulation. However, few works have been carried out for multi-objective combinatorial optimization MOCO especially in large-scale and service-oriented distributed simulation systems (SoDSSs). The existing algorithms for MOCO in SoDSSs are far from enough owing to their low accuracy or long decision time. To overcome this bottleneck, in this paper, a quantum multi-agent evolutionary algorithm (QMAEA), for addressing MOCO in large-scale SoDSSs is proposed. In QMAEA, the concept and characteristics of agent and quantum encoding are introduced for high intelligent searching. Each agent represented by a quantum bit, called a quantum agent (QAgent), is defined as a candidate solution for a MOCO problem, and each QAgent is assigned an energy, which denotes the fitness or objective function value of the candidate solution represented by it. Each QAgent is connected by four other QAgents nearby, and all QAgents are organized by an annular grid, called a multi-agent grid (MAG). In a MAG system, the population of QAgents can reproduce, perish, compete for survival, observe and communicate with the environment, and make all their decisions autonomously. Several operators, i.e. energy-evaluation-operator, competition-operator, crossover-operator, mutation-operator and trimming-operator, are designed to specify the evolvement of the MAG. The theory of predatory search strategy of animals is introduced in the evolution of QMAEA. Multiple evolutionary strategies, such as local-evolution-strategy, local-mutation-strategy and global-mutation-strategy are designed and used to balance the exploration (global search ability) and the exploitation (local search ability) of QMAEA. The framework and procedures of QMAEA are presented in detail. The simulation and comparison results demonstrate the proposed method is very effective and efficient for addressing MOCO in SoDSSs.
    Simulation 02/2014; 90(2):182-204. · 0.69 Impact Factor