... jamal.zemerly}@kustar.ac.ae. AbstractâTo provide elasticity for cloud-hosted applications, there is a need to specify the appropriate location of the resources.
Network-Aware Resource Allocation for Cloud Elastic Applications Fatima Mohammed AlQayedi, Khaled Salah, M. Jamal Zemerly Electrical and Computer Engineering Department Khalifa University of Science, Technology and Research, PO Box 573, Sharjah, United Arab Emirates {fatima.alqayedi, khaled.salah, jamal.zemerly}@kustar.ac.ae
provisioning happens when the demand for resources is lower than the available resources. It means, there are extra resources unused and the user still is paying for these unused resources. The second scenario is under provisioning where the demand is very high and there are not enough resources to handle heavy load from user side. These two scenarios affect the application performance in terms of response time and the user budget adversely. Therefore, there is a need to allocate the servers resources (computing and storage resources) dynamically according to the input load. This is called elastic resource provision (see Figure 1).
Abstract—To provide elasticity for cloud-hosted applications, there is a need to specify the appropriate location of the resources (computing and storage resources) while meeting SLA promises. This paper presents literature review, problem statement and research methodology.
Index Terms—Cloud computing, network-aware, elasticity, elastic applications, resource allocation, SLA and SLO. I. INTRODUCTION Cloud elastic application is an application which is running in cloud environment with elasticity support (i.e. banner, moodel, and blackboard). Cloud computing elasticity means the ability of the cloud provider to scale up and down the computing resources according to the fluctuation of the applications workload in order to meet Service Level Agreement (SLA) with the minimum cost. In this context, the elastic resource provision means the ability to configure and prepare the network to provide new service to the user. Provisioning must be achieved dynamically based on the allocation/de allocation of the Virtual Machines (VM). In another words, computing resources are scaled up/down depending on changing of the load over the time. Also attention should be given to the network bandwidth to maximize utilization of network links and computing resources and decrease network latency to users.
Figure 1 Imperfect capacity vs perfect capacity. adapted from [1]
III. LITERATURE REVIEW Although, there is steady increase in the elastic approaches found in the literature that present on-demand resource auto scaling, there is a shortcoming for the existing provisioning algorithms that they do not consider the network traffic required by each instance. There are numerous researches in resource allocation wither considering network issue or not.
II. PROBLEM STATEMENT AND POTENTIAL BENEFITS From user perspective, computing and storage resources should be available in infinite manner any time the user request them. So, the Cloud provider must have ability to auto scale the resources dynamically. The auto scaling of resources needs to know the appropriate physical machine for placement in order to meet SLA.
Mansoor and Lakshman [2] proposed a network aware resource allocation model in order to minimize the latency and cost of communication by minimizing the distance between user and the service vendor. Jung and Sim proposed a location aware model for resource allocation [3] to evaluate physical machines and network delay. A threshold is set to evaluate the PMs utilization to inform the provider if there is a need to migrate VM to another location.
Thus, meeting SLA promises requires from cloud provider to consider the place of VMs and network links capacity that affects the system performance in term of response time. Figure 1 illustrates the concept of elasticity by showing two main cases that may happen, while the applications are running in cloud environment, which are: resource over provisioning and resource under provisioning. Resource over
978-1-4799-2452-3/13/$31.00 ©2013 IEEE
88
However, Gurusamy et al. [4] proposed an interesting scheme where the VMs placement depends on the network bandwidth in order to reduce the cost of communication between Virtual Data Centers (VDCs) where the VMs allocated.
effectiveness of the proposed solution will be verified and compared with existing techniques using mathematical modeling and simulation tools.
CloudScale [5] predicts the demand of the resources with two error estimation approaches that depend on the true resource data demand. KingFisher [6] is a cost aware system that is based on migration and horizontal scale as elastic solutions. Kingfisher has the following advantages; (i) reduce the cost of the whole server types (small, medium, large) in order to reduce the provision cost. (ii) reduces the transition time, the time needed to add extra resources, by using the appropriate elastic decisions.
In this paper, we presented very recent and short literature survey on the existing proposed schemes that support elasticity in the cloud computing. We will propose and design a novel allocation scheme for VM instance requests in cloudhosted applications which takes into account realistic and dynamic parameters with the aim of yielding superior performance and efficiency over existing resource allocation schemes and verify, compare and evaluate the efficiency and performance of our proposed allocation schemes using mathematical modeling and simulation.
V. CONCLUSION
Han et al. [7] proposed an elastic approach for multi-tier cloud applications where the workload adaptive and cost of the infrastructure services are taken into account in order to maintain the quality of service and provide elastic management for the multi tier applications. Lightweight Resource Scaling [8] is extension of the work of [7] by using vertical and horizontal scaling techniques besides considering a cost effective approach to meet elasticity for cloud applications.
REFERENCES
The Dynamic service placement algorithm [9] monitors both network resources, computation and storage resources to increase the cloud elasticity by reducing bandwidth cost, increasing response time for the user and increasing the service availability. ConPaas [10] is a platform as a service environment that facilitates deploying applications in the cloud as services. The aim of ConPaas is to dynamically adapt the infrastructure capacity according to the workload fluctuation. Horizontal scale is the method of providing elasticity in ConPaas. IV. PROPOSED METHODOLOGY The aim of this research is to propose a novel networkaware resource allocation scheme for cloud elastic applications. Generally, the requests, that are entering cloud computing environment, are organized in a queue, therefore, the proposed scheme will be based on queuing theory of cloud environment where the requests are organized in a queue to server. The proposed scheme should take into account network resources beside servers’ resources while meeting SLA promises. In most of the proposed algorithms, VM instances are assumed to be fixed in compute and storage sizes.The proposed approach will overcome the shortcomings of the existing approaches in which VM instances are provisioned without considering the variability of compute, storage, and network resources required by each instance. The
[1]
Joe Weinman, "Time is Money: The Value of OnDemand",http://www.joeweinman.com/Resources/Joe_Weinman_Time _IMoney.pdf .
[2]
M. Alicherry and T. V. Lakshman, "Network aware resource allocation in distributed clouds", INFOCOM, 2012 Proceedings IEEE, Orlando, FL, pp. 963-971, 25-30 March 2012.
[3]
G. Jung and K. M. Sim, "Location-Aware Dynamic Resource Allocation Model for Cloud Computing Environment", International Conference on Information and Computer Applications (ICICA), IACSIT Press, Singapore, 2012.
[4]
M. Gurusamy, T. N. Le; D. M. Divakaran, "An integrated resource allocation scheme for multi-tenant data-center," 37th IEEE Conference on Local Computer Networks (LCN), Clearwater, FL, pp. 496-504, 2225 Oct. 2012.
[5]
Z.shen, S. Subbiah, X.Gu, and J.Wilkes, "Cloudscale: elastic resource scaling for multi-tenant cloud systems", Proc. 2nd symposium on cloud computing, ser.SOCC 2011. ACM, 2011, Portugal, pp. 1-14, October 26-28, 2011.
[6]
U. Sharma, P. Shenoy, S. Sahu, A. Shaikh, "A Cost-Aware Elasticity Provisioning System for the Cloud", 31st International Conference on Distributed Computing Systems (ICDCS), Minneapolis, MN, pp. 559570, 20-24 June 2011.
[7]
R. Han, M. M. Ghanem, L. Guo, Y. Guo, and M. Osmond, "Enabling cost-aware and adaptive elasticity of multi-tier cloud applications", Future Generation Computer Systems, Online June 2012, corrected proof in press.
[8]
R. Han, L. Guo, M. M. Ghanem, Y.Guo, "Lightweight Resource Scaling for Cloud Applications", 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), 2012, Ottawa, pp. 644-651, 13-16 May 2012.
[9]
M. Steiner, B. Gaglianello, V. Gurbani, V. Hilt, W. D. Roome, M. Scharf, and T. Voith, "Network-Aware Service Placement in a Distributed Cloud Environment", Bell Labs, Alcatel-Lucent, SIGCOMM 12, 13-17 August, Helsinki, Finland.
[10] G. Pierre, C. Stratan, "ConPaaS: A Platform for Hosting Elastic Cloud Applications", Internet Computing, IEEE , vol. 16, no. 5, pp.88-92, Sep.-Oct. 20 2012.
89
