Observation of Energy Efficiency in Green Cloud

Simulator
Mr. Anup Singh Kushwaha Dr.Bashir Alam Mrs.Gaganjot Kaur
Dept. of Computer Engineering Dept of Computer Engineering Dept of CST, Assistant Professor
Research Scholar, JMI DELHI Asst.Professor,JMI DELHI Manav Rachna University Faridabad
[email protected] [email protected]

Abstract — Green cloud is a highly developed packet-level iii. The prerequisites of the memory denominate the
simulator used to compute energy focusing on the energies superlative range of the factitious memory source that
consumed by various data centers including servers, network can be pre owned in aggregating [1].
switches. Green cloud simulators are helpful in generating The edge of the protocol of the cloud computing [2] virtual
solutions for arranging loads for different number of machines, technology maintains the capability of procurement of mode
maintaining and safeguarding the resources, developing the rules on demand on pay as you use basis.
required for delivery and also gaining results for network
Energy efficiency of clouds depends on various characteristics
switches.
including simulation time of various resources. Depending on
The paper presents the comparison of proper use of overall
the energy of various data centers, switches, servers and
energy along with depicting graph for total energy along with
energy of various resources. It shows the various impact of clouds the carbon efficiency is either increased or decreased.
energy by increasing number of resources using green clouds. The Green cloud simulator has its original simulation result
for 144 servers with 1 cloud users. The parameters that
Index Terms — Green Cloud, simulation, Scheduling, switches, describe the functionality of cloud environment are specified
servers, cloud users. in various Tcl files in Green cloud simulator that may be
further separated and certified positioned on the customer
I. INTRODUCTION fulfillments. For cloud environment one of the important
Green Cloud is a lined up simulator that exercises the existing factors to determine energy depends on the number of users as
Network Simulator 2 (NS2) athenaeums for effective per servers where in user consumes less power as per the
processing of workhorses in cloud computing. It engraves the server or data center load.
various subsistence of cloud such as servers, switches, links II. Diversity of Green Cloud Simulator
for delivery and their efficacy. It can be advantageous in The diversity of the GreenCloud is its design of
progressing explications for surveying and allotment of communication with various switches data centers servers and
various reserves, cataloging assignments for number of clients, cloud users. The actual profitable aggregate standards of NS2
developing the rules required for delivery and also gaining instruments adequate TCP/IP protocol reference model that
results for network switches. The data center enhancing may grants combination of various transmission protocols such as
be predetermined employing this tool. NS2 mainly practices IP, TCP and UDP with the simulation [3]. The construction of
two languages C++ and Otcl (Tool Command Language). The Green clouds is essentially prolonged as two tier and three tier
commands from TCL are usually passed to C++ using an architecture. Three tier architecture incorporates various core
interface TclCL. Green Cloud handles 80% of the coding in switches with access and aggregate forth with it where in
C++ (TclCL Classes) and remaining 20% coding resolves in servers are absolutely subordinate on the number of racks
Tcl scripts (Commands are sent from Tcl to C++). Green used. The Green cloud Simulator primarily exercises two
Cloud is advanced by University of Luxembourg and power saving modes Dynamic Voltage and Frequency Scaling
released under the General Public License (GPL). It and Dynamic Power Management that consecutively
incorporates the specified three simulation guidelines: influences the carbon efficiency of clouds.
i. Physical or Virtual infrastructure elucidates the
III. RELATED WORK
Statistics and enumerates the requisites of approaching The main focus of the related work and the proposed work is
loads with the consideration to virtual infrastructure. basically to cut down the overall energy cost of clouds be it
Consistently the amount should be between 0 and 1. The including or excluding the energy of cloud users and
amount adjoining to 0 epitomizes a trivial physical resources. As per the research of the work previously been
infrastructure, although the amount equivalent or higher than 1 done it is predicted that the expenditure of dynamizing and
would pervade the infrastructure. refrigerating reports for 53% of absolute operative cost of
ii. Simulation duration displays the ultimate interval granted virtual machines [4]. In 2006, virtual machine infrastructure in
for load processing, while the time limit of loads has result on the US depleted extra then 1.5% of the entire energy
the condominium observance of loads. Lengthy cutoffs grant engendered in that year, and the percentage is calculated to
heavier loads to be processed in the same direction on an increase 18% annually [5]. The architectonics planned by
individual machine or a virtual machine. Nathuji and Schwan [6] of energy management systems for
physical or virtual infrastructures spots the load to be branched

978-1-4673-8203-8/16/$31.00 2016
c IEEE 135
into local and global policies. At the local alignment the IV. PROPOSED WORK
machine authorizes guest operating system’s power As mentioned above the main aim of the proposed work of
management blueprints. Unification of VMs are organized by this paper is to decrease the energy efficiency. In this paper
global policies administering live migration to redistribute the work done is basically the comparison of energy by
VMs. Energy efficiency can also be increased by increasing increasing the resources and the cloud users and then
the resource utilization for the virtual machines. As in for analyzing the result for the same. The absolute assignment of
example the energy consumption of a virtual machine [7] is green cloud simulator is planted on the three tier architecture
shown in Figure1. The Figure shows the boot up (t=10 to with the adoption of DENS that is Data Center Energy
t=30), CPU burn (t=40 to t=100), shut down (t=110 to t=122) Efficient Network Aware scheduling algorithm Figure2.
of a virtual machine. It is also demonstrated that the start and
the closedown employs certainly limited energy. So we should
put virtual machine to sleep mode to save the power consumed
when it does nothing.

Figure2: Three Tier Architecture

The data center topology includes mainly core network on top

below of which is aggregation network having modules that
are linked with the core network and below which are the rack
hosts that are compacted as an Access Network and are also
linked with the Aggregation Network. All these topology
values actually holds important portion of energy
Figure 1: Life of a virtual machine: start, execute, and rest [7] consumption. The power and energy consumption are
generally dependent on the CPU utilization.
Considerable amount of energy depends upon switches which
actually forms support of interdependence which conveys task V. EXPERIMENTS AND RESULTS
requisition to enumerating machines for processing. The
comprehensive energy of machines is clinged on the type of As per the original work done by team of Green Clouds those
switch along with number of ports, port transmission rates and are Dzmitry Kliazovich, Pascal Bouvry of University of
cabling solutions. The energy expended by switch can be Luxemborg and Samee U.Khan from North Dakota State
observed as [8]: University, U.S.A the data center topology sets the value of
switches and of cloud users. Aggregation network is
Pswitch=Pchassis+nlinecards·Plinecard + σோ௜ୀ଴ ൌ nports·Pr (1)
dependent on the core network as they are linked so its value
are calculated from the core switch only that is Aggregation
Where,
network is exponential of multiple of two of the core switch
Plinecard - Power absorbed by an operating linecard
value been mentioned. Three tier architecture is also defined
Pchassis - Power absorbed by the switch base hardware,
in different levels where in values are computed for Three Tier
Pr - Power absorbed by an alive port (transmitter) running at
High Speed then for Three Tier Debug and then for Three Tier
the rater. In (1) only the last component, Pr, extents with a
Heterogeneous Small and Default values are been mentioned.
switch’s transmission rate. Both Pchasis and Plinecard do not
Taking into account the original values set for three tier high
propagates with the communication rate and it can be
speed [9]
prevented by shutting down the hardware including switch
network and core .This evidence restricts the assistance of any
Number of Rack Hosts – 3
rate adaptive scheme as the combined consumption of switch
Access Switches – 256
transceivers accounts for just 3–15% of switch’s total energy
Aggregation Network – expr(2* top Ncore)
consumption [8].
Core Switches – 2

136 2016 6th International Conference - Cloud System and Big Data Engineering (Confluence)
For three tier debug
Number of Rack Hosts – 48
Access Switches – 3
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

For three tier heterogeneous small

Number of Rack Hosts – 48
Access Switches – 3
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

Default
Number of Rack Hosts – 3
Access Switches – 64
Aggregation Network – expr(2* top Ncore)
Core Switches – 8

Number of racks computed are expr[top (Naccess) * top

(Ncore)]
Figure 4: Summary of Simulation Result
Number of cloud users -1

Number of servers computed are expr[top (Nracks) *top

(Nrackhosts)].
Figure 3 shows the topology of the data center architecture.
Figure 4 shows the summary of the simulation including
complete energy efficiency of the switches and servers. Figure
5 shows the statistics of data center load. Figure 6 shows the
statistics of virtual machines. Figure 7 shows the overall
energy consumption.

Figure 5: Data Center Statistics

Figure 3: Building Topology

Figure 6: Virtual Machine Statistics

2016 6th International Conference - Cloud System and Big Data Engineering (Confluence) 137
Figure 7: Energy Consumption Statistics
FIGURE 8: SUMMARY OF SIMULATION RESULT
Above was the simulation result of the original work and in
the proposed work we have doubled the cloud users taking all
the resources same as above

For three tier high speed

Number of Rack Hosts – 3
Access Switches – 256
Aggregation Network – expr(2* top Ncore)
Core Switches – 2

For three tier debug

Number of Rack Hosts – 48
Access Switches – 3
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

For three tier heterogeneous small Figure 9: Data Center Statistics

Number of Rack Hosts – 48
Access Switches – 3
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

Default
Number of Rack Hosts – 3
Access Switches – 64
Aggregation Network – expr(2* top Ncore)
Core Switches – 8
Number of racks computed are expr[top (Naccess) * top
(Ncore)]

Number of cloud users -2

Number of servers computed are expr[top (Nracks) *top
(Nrackhosts)].
Figure 8 shows the summary of the simulation including
complete energy efficiency of the switches and servers. Figure
9 shows the statistics of data center load. Figure 10 shows the
statistics of virtual machines. Figure 11 shows the overall
energy consumption.
Figure 10: Virtual Machine Statistics

138 2016 6th International Conference - Cloud System and Big Data Engineering (Confluence)
Figure 11: Energy Consumption Statistics

Now we have reduced the sever from 144 to 96 by reducing

access switches to 2 from 3 for three tier debug and
heterogeneous small
For three tier high speed
Number of Rack Hosts – 3 FIGURE 12: SUMMARY OF SIMULATION RESULT
Access Switches – 256
Aggregation Network – expr(2* top Ncore)
Core Switches – 2

For three tier debug

Number of Rack Hosts – 48
Access Switches – 2
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

For three tier heterogeneous small

Number of Rack Hosts – 48
Access Switches – 2
Aggregation Network – expr(2* top Ncore)
Core Switches – 1

Default
Number of Rack Hosts – 3
Access Switches – 64
Aggregation Network – expr(2* top Ncore) Figure 13: Data Center Statistics
Core Switches – 8

Number of racks computed are expr[top (Naccess) * top

(Ncore)]

Number of cloud users -2

Number of servers computed are expr[top (Nracks) *top
(Nrackhosts)].
Figure 12 shows the summary of the simulation including
complete energy efficiency of the switches and servers. Figure
13 shows the statistics of data center load. Figure 14 shows the
statistics of virtual machines. Figure 15 shows the overall
energy consumption.

Figure 14: Virtual Machine Statistics

2016 6th International Conference - Cloud System and Big Data Engineering (Confluence) 139
 [3] Gaganjot Kaur, “Study of Comparison of Various Cloud Simulators”,
Proceedings at 2nd National Conference on Emerging Trends in Intelligent
Computing & Communication organized by Galgotia College of Engineering
& Technology-, held on,26, 27,April 2013.

[4] J. Hamilton, ʊCooperative expendable micro-slice servers (CEMS):

low cost, low power servers for Internet-scale services, In the
Proceeding of CIDR, 2009

[5] B. Li, ʊEnaCloud: an energy-saving application live placement

approach for cloud computing environments, In the Proceeding of
international conference on cloud computing, 2009.

[6] R. Nathuji and K. Schwan, “Virtualpower: Coordinated power

management in virtualized enterprise systems,”ACM SIGOPS Operating
Systems Review, vol. 41, no. 6, pp. 265–278, 2007.

[7] L. Lefevre and A. C. Orgerie, ʊDesigning and evaluating an

energy efficient Cloud, The Journal of Supercomputing, Springer, Vol. 51,
Issue no. 3, page no.
352-373, March 2010.

[8] L. Liu, H. Wang, X. Liu, X. Jin, W. B. He, Q. B. Wang, and Y. Chen,

“Greencloud: a new architecture for
Figure 15: Energy Consumption Statistics green data center,” in ICAC-INDST ’09: Proceedings of the 6th international
conference industry session on Autonomic computing and communications
industry session. New York, NY, USA: ACM, 2009, pp. 29–38.
VI. CONCLUSION
[9] DENS: Data Center Energy-Efficient Network-Aware Scheduling
This paper basically works on the comparison of result of IEEE/ACM International Conference on Green Computing and
energy of the clouds by varying the number of resources and Communications (GreenCom), Hangzhou, China, 2010 -[Best Paper Award]
Authors: D. Kliazovich, P. Bouvry, and Samee U. Khan
cloud users. As per the comparative study when we doubled
the number of cloud users to the 100 % so accordingly the
energy efficiency should also be doubled that is to be
increased 100% only but instead it is just increased 46%. We
can check through values as well when cloud user = 1 the total
energy efficiency was 306.2 W*h and when we doubled the
cloud user = 2 the total energy efficiency came up to 447.5
W*h. Similarly the evaluation was made when the number of
resources that is the server were decreased and cloud users
were doubled so the effect on entire energy came up to 351
W* h which means as compared to the original values
doubling the number of users and decreasing the servers the
effect of energy is just 14% and when compared to our second
case that is with just doubling the cloud users keeping number
of resources just same the increase in energy is just 27%. So as
per our work be it if we double the cloud users our energy of
simulator is still efficient. We have presented a comparative
study for total energy efficiency including the energy of core,
access and aggregate switches, servers.
Future Research work includes better alternative to data
center energy efficient aware scheduling algorithm in Green
Cloud Simulator so as to reduce more energy consumption and
thus reducing carbon footprints with increasing the cloud users
or simulation time.

REFERENCES
[1] Greencloud Simulator User Manual. University of Luxembourg. The
Green Cloud Team December 19, 2013.

[2] R. Buyya, C. S. Yeo, and S. Venugopal, “Market-orientedcloud

computing: Vision, hype, and reality for delivering it services as computing
utilities,” in Proceedings of the 10th IEEE International Conference on High
Performance Computing and Communications (HPCC’08). IEEE CS Press,
Los Alamitos, CA, USA, 2008.

140 2016 6th International Conference - Cloud System and Big Data Engineering (Confluence)