Professional Documents
Culture Documents
Abstract
Cloud computing refers to applications and services that run on a distributed network using virtualized resources and accessed
through web as utility. Resource provisioning, policies allow efficient sharing of resources available in a data center, and these
policies help to evaluate and enhance the cloud performance. Resource provisioning that maintains the quality of service with
best resource utilization is a challenge. The template based resource provisioning and utilization method (TBRP) overcomes the
problem of over-provisioning and under-provisioning of resources by designing appropriate templates with QoS parameters
which avoid SLA violations. The results show that in TBRP method one need to design variable template for better performance
and use depending upon the capacity of Data-Center.
Keywords: Service Level Agreement, Quality of Service, Virtual Machines, Resource Provisioning.
1. INTRODUCTION
The implementation and escalation of cloud computing represent a paradigm shift by outsourcing IT and computational
needs. Cloud computing is an abstraction based on the notion of pooling physical resources and presenting them as a
virtual resources to access information anytime anywhere. It is evolution of a variety of technologies that bundles
together to provide IT infrastructure as per organization’s needs [13], [14]. Cloud provider offers software, platform
and infrastructure as a service to cloud users. Cloud users and Cloud service providers have to agree upon Service level
attributes which are specified in the Service Level agreement (SLA). The terms in SLA are responsible for maintaining
the Quality of Service (QoS) which can be monitored, measured and controlled [2]. The consumer of the service may
provide one or more service level objectives depending on specific application requirements but the cloud provider has
to translate them into low level (uptime, downtime, CPU utilization, etc.) technical attributes that can be monitored and
controlled to achieve the higher level (availability, reliability, etc.) Objectives [12]. The main objective of a cloud
service provider is to minimize the cost yet maintaining customer satisfaction.
Load balancing at Resource provisioning involves the distribution of resources to different cloud users without
increasing unused capacities and yet maintaining required Quality of service [1], [4], [8], [10] [17]. Resource allocation
is a main issue in cloud computing environment. Scheduling task, computational performance, reallocation, response
time and cost efficiency are variant level of issues in resource allocation. Resource allocation is the process of providing
services and storage space to the particular task given by the users. One of the very important issue in resource
allocation is capacity planning. Capacity planning seeks to match demand to available resource so as to accommodate
the workload where resources and tasks are diverse. In case of some tasks there is a high demand for the CPU, while in
some other case more storage is important issue. One main focus is to reduce time during scheduling and improving
performance, w.r.t response time, optimal span and completion time.
Template-Based Resource Provisioning (TBRP) method [5] for resource provisioning and optimum utilization of
idle capacity without breaking the SLA on variation of workload, need to be measured on different capacity Data-
Center works with fixed template or variable templates is the objective of this research.
Objectives of Research
• To study whether different configuration of basic VMs give better resource utilization in Data-Center.
• To find whether variable Template improve resource utilization in different capacity Data-Centers.
To test the objectives of this research, TBRP is applied on three different capacity Data-Center with fixed or
variable templates for comparison and analysis. The simulation experiments are carried out to compare completion
time as one of QoS parameter in different workload scenarios for different capacity Data-Center for fixed and variable
templates.
Capacity of Data-Centers vary from Data-Center to Data-Center. In CloudSim environment, Datacenter consists of
fixed or varied configuration of hosts (servers). In TBRP method, entire resource is divided into a set of VM types that
is small, medium and large which are priced according to their characteristics such as MIPS rating, RAM, and storage
etc. [5], [6], [18]. For the experiment three Data-Center are considered with different MIPs rating for small, medium
and large VMs as presented in Table-2.
Table-2: Data-Center with MIPs rating of VMs
The number of VMs of each type is fixed for each Data-Center. The overhead costs vary from Data-Center to Data-
Center and the pricing strategy for VM templates need to provide for these issues.
The customer must choose between the available templates that satisfy their QoS requirements or negotiate for a
custom-made template. The next step is to design templates with appropriate SLA clauses by simulating different
workload situations.
Table-9: Available template with different MIPs capacity for three Data-Center
Table-11: Completion time range of SLA clause for VM templates in DC-1, DC-2 and DC-3
Low workload
1.0 0.33 0.33 0.33
2 100 200 0.66 1 0.5 0.83 1 2 0.5
2.0 0.66 0.66 0.66
4 100 400 0.66 1 0.75 1.5 2 3.99 0.75
2.98 0.99 0.99 0.99
6 100 600 1 1 1 2 2.98 5.96 1
3.98 1.33 1.33 1.33
8 100 800 1.33 1.5 1.24 2.74 3.99 7.96 1.24
5.0 1.66 1.66 1.66
10 100 1000 1.99 2 1.74 3.38 5.17 10 1.74
Medium workload
3.75 5.0 3.33 3.33 3.33
2 1000 2000 6.66 10 5.0 8.33 10 8.33
7.5 7.5 6.66 6.66 6.66
4 1000 4000 6.67 10 7.5 9.16 20 10.83
9.99 10.0 9.99 9.99 9.99
6 1000 6000 10 10 10 10.55 29.97 16.1
13.43 13.74 13.33 13.33 13.33
8 1000 8000 13.33 15 13.74 14.57 39.99 22.07
16.99 16.99 16.66 16.66 16.66
10 1000 10000 16.66 18 16.99 19.6 49.99 28.32
Large workload
37.5 50.0 33.33 33.33 33.33
2 10000 20000 66.66 100 50 83.33 50 58.33
75.0 75.0 66.66 66.66 66.66
4 10000 40000 66.66 100 75 91.66 75 108.32
100.0 100.0 99.99 99.99 99.99
6 10000 60000 111.11 100 100 105.55 100 116.66
134.37 137.49 99.99 99.99 99.99
8 10000 80000 133.33 150 137.49 141.66 150 154.16
172.3 199.9 169.9 173.3 169.9 176.6 169.99 169.99 166.66 166.66 166.66
10 10000 100000 3 9 9 2 9 5
The completion time range of SLA clauses on above defined VM-Templates are presented in Table-11 and are
extracted from Table-10 for different workload scenarios.
Table-10: Template execution with 600 MIPs capacity for three Data-Center
Figure-1: Template completion time in three Data-Center with 600 MIPs capacity for all workload
Figure-2: Template completion time at DC-1, DC-2 and DC-3 for low workload
Observation:
Figure-1 presents template execution with 600 MIPs capacity at three Data-Center for low, medium and large
workload. It is observed from Figure-2 that for low workload templates DC3T001, DC2T020, DC1040 performed
uniformly with good results whereas DC1T121 performs poorly as compared to all other templates. Similar results are
observed for medium and large workloads. Table-12 presents completion time values of 900 MIPs capacity templates in
DC-2 and DC-3.
Table-12: Template completion time for DC-2 and DC-3 with 900 MIPs capacity
Figure-3: Template completion time at DC-2 and DC-3 for all workload
Figure-4: Template completion time at DC-2 and DC-3 for low workload
Figure-5: Template completion time at DC-2 and DC-3 for medium workload
Figure-6: Template completion time at DC-2 and DC-3 for large workload
Observations: Figure-3 presents template execution with 900 MIPs capacity at two Data-Center for low, medium and
large workload. For low workload performance of templates DC3T020 and DC3T300 is better. For templates DC2310
and DC2030 performance is good, whereas template DC2T111 performs poorly as presented in Figure-4. It is observed
from Figure-5 that for medium workload templates DC3T020, DC3T101, DC3T300 and DC2T310 performs uniformly,
whereas template DC2T111 performs poorly. For large workload all templates performs uniformly with better result for
template DC3T020 as presented in Figure-6.
Discussion: In this experiments even though the capacity of DC-1, DC-2 and DC-3 is different, fixed VM
configuration of VM template results into same performance with poor resource utilization. In case of variability in
configuration of basic VMs improve performance as well as improves resource utilization
7. CONCLUSION
Template based resource provisioning and utilization method and procedure overcomes the problem of over-provision
and under-provision of resources at Data-Center without effecting QoS and meeting SLA. Capacity of Data-Center
plays important role in designing the template If Data-Center is of low capacity and grow to high capacity same TBRP
method can be applied. Result shows that if configuration of basic VMs in templates are fixed for different capacity
Data-Center there is no change in performance but resource utilization is poor. In case of variability in configuration of
basic VMs improve performance as well as improves resource utilization. In TBRP method one need to design variable
template for better performance and utilization depending upon capacity of Data-Center.
REFERENCES
[1] Bianco, P., Lewis, G. A., & Merson, P. (2008). Service level agreements in service-oriented architecture
environments (No. CMU/SEI-2008-TN-021). Carnegie-Mellon Univ Pittsburgh Pa Software Engineering Inst.
[2] Buyya, R., Abramson, D., & Giddy, J. (2000, June). An Economy Driven Resource Management Architecture for
Global Computational Power Grids. In PDPTA (pp. 26-29).
[3] Buyya, R., Yeo, C. S., Venugopal, S., Broberg, J., &Brandic, I. (2009). Cloud computing and emerging IT
platforms: Vision, hype, and reality for delivering computing as the 5th utility. Future Generation computer
systems, 25(6), 599-616.
[4] Byun, E. K., Kee, Y. S., Kim, J. S., &Maeng, S. (2011). Cost optimized provisioning of elastic resources for
application workflows. Future Generation Computer Systems, 27(8), 1011-1026.
[5] Chowhan, S. S., Shirwaikar, S., & Kumar, A. (2019). Template-Based Efficient Resource Provisioning and
Utilization in Cloud Data-Center, International Journal of Computer Sciences and Engineering, Vol.7, Issue.1,
pp.463-477.
[6] Garg, S. K., Gopalaiyengar, S. K., &Buyya, R. (2011, October). SLA-based resource provisioning for
heterogeneous workloads in a virtualized cloud datacenter. In International conference on Algorithms and
architectures for parallel processing (pp. 371-384). Springer, Berlin, Heidelberg.
[7] Greenberg, A., Hamilton, J., Maltz, D. A., & Patel, P. (2008). The cost of a cloud: research problems in data
center networks. ACM SIGCOMM computer communication review, 39(1), 68-73.
[8] John, M., Gurpreet, S., Steven, W., Venticinque, S., Massimiliano, R., David, H., & Ryan, K. (2012). Practical
Guide to Cloud Service Level Agreements.
[9] Kremer, J. (2010). Cloud Computing and Virtualization. White paper on virtualization.
[10] Liu, F., Tong, J., Mao, J., Bohn, R., Messina, J., Badger, L., & Leaf, D. (2011). NIST cloud computing reference
architecture. NIST special publication, 500(2011), 1-28.Malhotra, L., Agarwal, D., &Jaiswal, A. (2014).
Virtualization in cloud computing. J Inform Tech SoftwEng, 4(2), 136.
[11] Nurmi, D., Wolski, R., Grzegorczyk, C., Obertelli, G., Soman, S., Youseff, L., &Zagorodnov, D. (2009, May).
The eucalyptus open-source cloud-computing system. In Proceedings of the 2009 9th IEEE/ACM International
Symposium on Cluster Computing and the Grid (pp. 124-131). IEEE Computer Society.
[12] Radojević, B., & Žagar, M. (2011, May). Analysis of issues with load balancing algorithms in hosted (cloud)
environments. In 2011 Proceedings of the 34th International Convention MIPRO (pp. 416-420). IEEE.
[13] Shawish, A., &Salama, M. (2014). Cloud computing: paradigms and technologies. In Inter-cooperative collective
intelligence: Techniques and applications (pp. 39-67). Springer Berlin Heidelberg.
[14] Sosinsky, B. (2010). Cloud computing bible (Vol. 762). John Wiley & Sons.
[15] Tsakalozos, K., Kllapi, H., Sitaridi, E., Roussopoulos, M., Paparas, D., & Delis, A. (2011, April). Flexible use of
cloud resources through profit maximization and price discrimination. In 2011 IEEE 27th International
Conference on Data Engineering (pp. 75-86). IEEE.
[16] Vinothina, V., Sridaran, R., & Ganapathi, P. (2012). A survey on resource allocation strategies in cloud
computing. International Journal of Advanced Computer Science and Applications, 3(6), 97-104.
[17] Wu, L., &Buyya, R. (2012). Service level agreement (sla) in utility computing systems. IGI Global, 15.
[18] Wu, L., Garg, S. K., & Buyya, R. (2011, May). SLA-based resource allocation for software as a service provider
(SaaS) in cloud computing environments. In Proceedings of the 2011 11th IEEE/ACM International Symposium
on Cluster, Cloud and Grid Computing (pp. 195-204). IEEE Computer Society.
[19] Zhang, X., Wu, T., Chen, M., Wei, T., Zhou, J., Hu, S., & Buyya, R. (2019). Energy-aware virtual machine
allocation for cloud with resource reservation. Journal of Systems and Software, 147, 147-161.
AUTHOR
Ms. Seema Chowhan is working as a faculty and head in subject of computer science in Baburaoji Gholap
College Pune, India affiliated to Savitribai Phule Pune University, Pune. She has 18+ years of experience in
teaching UG and PG courses. She has completed M.Phil (CS).Her research interests include Cloud Computing
and Networking.
Dr. Ajay Kumar experience covers more than 26 years of teaching and 6 years of Industrial experience as IT
Technical Director and Senior Software project manager. He has an outstanding academic career completed
B.Sc. App. Sc. (Electrical) in 1988, M.Sc. App.Sc. (Computer Science-Engineering and Technology) in1992 and
PhD in1995. Presently, working as Director at JSPMs Jayawant Technical Campus, Pune (Affiliated to Pune
University). His research areas are Computer Networks, Wireless and Mobile Computing, Cloud computing,
Information and Network Security. There are 74 publications at National and International Journals and
Conferences and also worked as expert, appointed by C-DAC to find Patent-ability of Patent Applications in ICT area. Six
commercial projects are completed by him for various companies/ Institutions. He holds variety of imperative position like
Examiner, Member of Board of Studies for Computer and IT, Expert at UGC.
Dr. Shailaja Shirwaikar has a Ph. D. in Mathematics of Mumbai University, India and worked as Associate
Professor at Department of Computer Science, Nowrosjee Wadia College affiliated to Savitribai Phule Pune
University, Pune for last 27 years. Her research interests include Soft Computing, Big Data Analytics, Software
Engineering and Cloud Computing.