Ansuyia Makroo ¹, Deepak Dahiya ¹

Affiliations
1 School of Engineering and Technology, Ansal University, Gurgaon - 122003, Haryana, IN

Abstract

Background/Objectives: One of the major challenges of the multitenant cloud model is performance unpredictability because of resource contention. The objective of this paper is to propose an approach to deal with noisy neighbours in a shared cloud infrastructure and reduce their affect on the other tenant applications. Methods/Statistical Analysis: Multiple tenant applications are deployed on cloud VMs which share the underlying system resources. Noisy neighbour applications take up more resources and leave other applications in turmoil state that leads to lack of predictability of the performance of other applications. The proposed system actively monitors the resource consumption of the tenant applications based on some identified parameters. Findings: Monitoring the resource consumption of the applications helps categorize the resource greedy applications as noisy neighbours. The other tenant applications which do not get their fair share of the resources are victim applications. Once the noisy neighbours have been identified, the next step is to deal with noisy neighbours by either migration of victim applications on a separate node on the cloud or borrowing resources from other nodes or implementing a quota system for resource allocation in cloud. Applications/Improvements: The pragmatic study of the behaviour of tenant applications based on their resource consumption pattern would help researchers to better focus on techniques to improve the Quality of Service of a cloud.

Keywords

Cloud Computing, Multitenant, Noisy neighbour, Quality of Service, Virtual Machine

Full Text

   

Ansuyia Makroo ¹, Deepak Dahiya ¹

Affiliations
1 School of Engineering and Technology, Ansal University, Gurgaon - 122003, Haryana, IN

Abstract

Objectives: One of the essential requirements for improving Quality of Service of Cloud is to design load balancing algorithms that equally spread the load among the VMs such that neither of them is overloaded at any given point of time while ensuring that each of the VM is optimally utilised. The objective of this paper is to propose an efficient VM load balancing policy called Timestamp based Stateful Throttled VM load balancing algorithm. Methods/Statistical Analysis: The proposed algorithm deals with the space performance constraints of the existing Stateful Throttled VM Load balancing algorithm. It is based on the principle of locality of reference and deletes the old state information to reclaim space. The authors have carried out an experimental analysis to compare the proposed algorithm with the existing algorithms. Findings: The comparative analysis shows that the algorithm behaves in the same manner as Stateful Throttled Algorithm while taking into consideration its space performance limitations. The authors have observed better performance in terms of response time and data center processing time for Timestamp based Stateful Throttled VM load balancing algorithm than Throttled VM load balancing algorithm in case of spatially distributed VMs. Application/Improvements: The salient features of the proposed algorithm are better response time and data center processing time in case of spatially distributed data centers in cloud.

Keywords

Cloud Computing, Data Center, Load Balancing, Virtualization, Virtual Machine.

Full Text