Energy efficient task scheduling algorithms for cloud data centers

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 03 | Mar-2014, Available @ http://www.ijret.org 322
ENERGY-EFFICIENT TASK SCHEDULING ALGORITHMS FOR CLOUD
DATA CENTERS
T. Jenifer Nirubah1
, Rose Rani John2
1
Post-Graduate Student, Department of Computer Science and Engineering, Karunya University, Tamil Nadu, India
2
Assistant Professor, Department of Computer Science and Engineering, Karunya University, Tamil Nadu, India
Abstract
Cloud computing is a modern technology which contains a network of systems that form a cloud. Energy conservation is one of the
major concern in cloud computing. Large amount of energy is wasted by the computers and other devices and the carbon dioxide gas
is released into the atmosphere polluting the environment. Green computing is an emerging technology which focuses on preserving
the environment by reducing various kinds of pollutions. Pollutions include excessive emission of greenhouse gas, disposal of e-waste
and so on leading to greenhouse effect. So pollution needs to be reduced by lowering the energy usage. By doing this, utilization of
resources should not be reduced. With less usage of energy, maximum resource utilization should be possible. For this purpose, many
green task scheduling algorithms are used so that the energy consumption can be minimized in servers of cloud data centers. In this
paper, ESF-ES algorithm is developed which focuses on minimizing energy consumption by minimizing the number of servers used.
The comparison is made with hybrid algorithms and most-efficient-server first scheme.
Keywords: Cloud computing, Green computing, Energy-efficiency, Green data centers and Task scheduling.
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
Cloud computing is a computing model which is developed
based on other computing models like grid computing,
distributed computing, parallel computing, cluster computing,
etc [1]. The main goal of cloud computing is to use the
resources in an efficient way and also to gain large profit [2].
The principle behind cloud computing is pay for what is used.
In cloud environment, the users can share data and resources
among them. The major services provided by the cloud are
Saas (Software as a Service), Iaas (Infrastructure as a Service),
Paas (Platform as a Service). The datacenters in Clouds has
increased the usage of computers in each year, which
increases the consumption of energy and also negative impact
on the environment [3]. While speaking about energy
consumption, green computing technology must be included
as it deals with reducing pollution by conserving energy.
Green computing can refer to energy-efficient personal
computers. The purposes of energy consumption reduction
are: minimizing performance losses, achieving target battery
lifetime, satisfying performance requirements, minimizing
power consumption, minimizing the CO2 emissions,
maximizing the profit, maximizing resource utilization [4].
As the number of users in cloud is increasing, the tasks to be
scheduled also increases. The performance of the cloud is
dependent on the task scheduling algorithms that is being used
[5]. Cloud servers are normally implemented in one or more
data centers. The major component of a datacenter in energy
consumption is the servers. The usage of servers in an
inefficient manner leads to more energy being consumed [6].
In data centers, the reduction in energy consumption can be
done by reducing the number of active servers. Initially more
servers are provided to meet the increase in demand. So many
task scheduling algorithms have been developed and proposed
so far with the aim of reducing the energy consumption. All
the task scheduling algorithms are developed in such a way
that the incoming tasks will be scheduled based on a
procedure given by the algorithm. Also the condition that the
tasks should meet with the deadline is essential. In [7], many
green task scheduling algorithms are developed with
continuous and discrete speeds on heterogeneous computers.
Such algorithms are needed so that the emission of green
house gas being emitted in the atmosphere can be reduced by
lowering the energy consumption. Data centers have many
servers. Energy consumed by the servers is increasing to a
greater extent. A task scheduling algorithm is needed for
minimizing energy consumption in cloud computing data
centers. With the help of the task scheduling algorithm, the
tasks can be scheduled to a minimum number of servers by
maintaining the task response time constraints. The number of
servers that can be active in order to meet the time constraints
must be taken into consideration. Data centers that enhance
the green computing technology can be called as Green Data
Centers. The main aim is to minimize the allocated number of
servers and thus to save the energy. Data center comprises of
many computer systems and components for communication,
processing and storage. A data center size may vary like
occupying one room of a building, one or more floors or an
entire building [8]. In order to maintain the energy efficiency
in data centers, various factors are needed to be considered. If
there are any servers that are not in use, then they can be

__________________________________________________________________________________________
turned off. Some servers are under-utilized and they run at a
maximum of 15 percent. Hence tasks in such servers can be
given to other high efficient servers which are having less
space and so the former servers can also be turned off [9]. The
servers which process huge number of tasks will be more
efficient in terms of utilization rate. Inefficient power supplies
can also result in wasting half of the power before it goes to
the server. The tasks given to the servers should be scheduled
within the particular processing time. The data center consists
of many switches and routers which enables communication
between the servers. In data centers, the task scheduling and
allocation to servers is done with the help of a scheduling
algorithm. Users send their request to the data center for
computing whatever task they want to get executed. The task
may be reading file contents, updating data, uploading files,
downloading software, etc. The data center will classify the
tasks based on the scheduling algorithm that is used and then
allocates the task to one of the available servers. The servers
that are active can be reduced in order to conserve energy.
Energy consumption can be maintained by making the active
servers to continue their work. ESF-ES algorithm is developed
by using the parameters from hybrid algorithms [7] and most-
efficient-server-first scheme algorithm [6].
2. RELATED WORK
A cloud computing system contains large number of
heterogeneous servers which execute tasks that are assigned to
them and consumes large amount of power. This is due to the
poor task assignment optimization. When the power
consumption is high, then the energy emission is high and the
carbon dioxide gas also will be emitted more into the
environment causing pollution. Therefore it is necessary to
develop efficient green computing algorithms to lower the
energy consumption on heterogeneous cloud servers. In [10],
some of the green task scheduling algorithms is used to
minimize the energy consumption in heterogeneous cloud
servers. In [11], green task scheduling algorithms with speed
adjustment and maximum speeds are used to reduce energy
consumption in heterogeneous computers and to finish the
tasks before deadline. In [7], several green task scheduling
algorithms are used in which some algorithms work for
continuous speeds and other algorithms for discrete speeds.
All these algorithms focus on minimizing the consumption of
energy and reducing the emission of carbon dioxide gas into
the atmosphere. A green scheduling algorithm is given in [3]
which use a neural network predictor to predict the future load
demand and the focus is on energy shortage and global climate
change problems. In [6], most-efficient-server first scheme is
used where energy consumption of servers in data centers is
calculated. The study of efficient task scheduling in order to
minimize the data center server energy consumption is
focused which can be achieved by reducing the number of
servers. In [12], various power-aware algorithms are used with
dynamic voltage and speed which are applicable to task
scheduling in real-time multiprocessing systems and
communication environments. Task scheduling using genetic
algorithm is discussed in [1,13] where the concern is about
efficient allocation and scheduling of tasks. Slack reclamation
is the execution of individual tasks with slacks. This concept is
used by [14,15] to focus on the energy consumption problems
in parallel and distributed computing systems with DVFS
method. The problem of power consumption in high
performance processors is addressed in[16] which also uses
slack reclamation. In [17], the need for scheduling algorithms
in order to minimize the wasted server energy is considered.
Energy-conscious scheduling algorithms are used to minimize
energy consumption in servers of data centers [18]. Power
consumption problem is considered in [19] and is optimized
through priority-based preemptive scheduling algorithm.
Power-aware scheduling algorithms with deadline constraints
for heterogeneous systems are used in [20] to meet the
deadline constraints in high performance computing
applications. Stochastic Max-Min and Min-Min algorithms are
used to solve the problem of scheduling multiple, divisible
independent tasks on a heterogeneous distributed computing
system [21]. A task clustering method is used in [22] to solve
the problem of task assignment in heterogeneous computing
systems and the tasks are assigned to the processors in such a
way that the communication cost and total execution cost are
minimized.
3. ESF-ES ALGORITHM
ESF_ES algorithm is developed by combining the hybrid
algorithm and most-efficient-server first scheme and is
implemented in this paper. Set of tasks and servers are taken
as input. The scheduling of tasks to the servers and the data
center server energy consumption is given as output of the
algorithm. The users will request for computing various types
of tasks. Each task may fall under a particular task type like
reading file contents, updating data, uploading files,
downloading software, etc. Based on the type of task selected,
the processing time vary. The number of instruction in each
task is obtained. Energy slope is calculated for each task of
different types in each server with the help of processing time.
Energy consumption is calculated by using the number of
instructions and the energy slope. Task allocation is done in
such a way that most-efficient-server gets the tasks first.
Number of active servers among the set of available servers is
reduced. The algorithm follows a greedy approach.
ESF-ES algorithm
Input: Set of tasks and servers.
Output: Scheduling of tasks to servers and data center server
energy consumption.
Get the number of tasks, number of available servers, task
type of each task and the number of instruction of each task.
a=b=1;
for each task x of type i do
for each server k do

__________________________________________________________________________________________
Calculate
1
,, ][ 
 i
kiiki ptC 

Calculate server energy consumption
kikiRkiE ,,),( 
if E(i,k) <= E(a,b) then
a=i, b=k
end
end
end
Output the task allocation to the servers and the total energy
consumption by the servers.
kipt , is the processing time for task of type „i‟ at server „k‟,
kiR , is the number of instructions of each task and ki, is the
energy slope of task „i‟ in server „‟k‟. This algorithm is helpful
in completing the sequential tasks on the heterogeneous
servers in order to minimize the consumption of energy in a
cloud computing scenario. Heterogeneous tasks are handled in
an efficient way. As like the algorithm in [6], this algorithm
also will use only the servers that are necessary. Additional
servers are taken only if the current servers are not adequate.
4. SIMULATION RESULTS AND
PERFORMANCE ANALYSIS
The simulation of ESF-ES green task scheduling algorithm for
a data center with heterogeneous tasks is done using java and
so that the server energy consumption can be reduced to a
certain extent. The reason for calling as green task scheduling
algorithm is that when the energy consumption is reduced, the
carbon dioxide gas emission can also be reduced resulting in
the enhancement of green computing technology. An energy-
efficient task scheduling algorithm namely ESF-ES is
implemented in this paper. Input is the number of servers
which handles heterogeneous tasks. The number of servers is
taken as 5. Four different task types are considered namely
reading file contents, updating data, uploading files and
downloading software. The algorithm is implemented in Java.
The processing time (ms) taken for each task varies depending
upon their task types. For every simulation, the processing
time (pt) and the number of instructions for each task (R) are
randomly generated. The processing time for the task type
reading file contents ranges between 1000 ≤ pt ≤ 1200; for
updating data, 1300 ≤ pt ≤ 1500; for uploading files, 1600 ≤ pt
≤ 1900; for downloading software, 2000 ≤ pt ≤ 2500. The
number of instruction for each task varies 1000 ≤ R ≤ 9000,
1000 ≤ iC ≤ 1500 and 3 ≤ i ≤ 4. The comparison is made
between three algorithms namely hybrid algorithm, most-
efficient-server first scheme and ESF-ES algorithm. The
energy consumption of data center servers is measured in
joules (J) and the processing time is measured in milli seconds
(ms). The processing time is used for calculating the energy
slope which in turn used in the calculation for energy
consumption.
Chart -1: Comparison based on number of tasks
Chart -2: Comparison based on number of servers
Chart -3: No. of active servers vs. no. of tasks
0.00E+00
5.00E+08
1.00E+09
1.50E+09
2.00E+09
2.50E+09
10 11 12 13 14
Energyconsumption(J)
No. of tasks
Hybrid
Most-
efficient-
server first
scheme
ESF-ES
0.00E+00
2.00E+08
4.00E+08
6.00E+08
8.00E+08
1.00E+09
1.20E+09
1.40E+09
1.60E+09
1.80E+09
2.00E+09
5 6 7 8 9
Energyconsumption(J)
No. of servers
Hybrid
Most-
efficient-
server first
scheme
ESF-ES
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
13 14 15 16
No.ofactiveservers
No. of tasks
Hybrid
Most-
efficient-
server first
scheme
ESF-ES

__________________________________________________________________________________________
In the first two graphs, the server energy consumption is
shown where the number of servers is fixed to 5 and the
number of tasks is varied from 10 to 14 in the first graph. As
the number of tasks to scheduled increases, the energy
consumption also increases. The hybrid algorithm is
consuming more energy than the other two algorithms. New
algorithm is better in energy conservation when compared to
hybrid algorithm. But most-efficient-server first scheme still
proved to be the best among the three algorithms. In the
second graph, the number of tasks is fixed to 12 and the
number of servers varies from 5 to 9. As the number of servers
increases, the energy consumption decreases. Again most-
efficient-server first scheme is better than the newly developed
algorithm and the hybrid algorithm. The third graph shows the
number of servers that are used when the number of tasks is
changed.
5. CONCLUSIONS
The various energy-efficient task scheduling algorithms are
studied. The ESF-ES algorithm is implemented which uses a
greedy task scheduling scheme and compared with the most-
efficient-server first scheme algorithm and the hybrid
algorithms. In the hybrid algorithm, only the number of tasks
is considered. In the most-efficient-server first scheme and the
ESF-ES algorithm, different types of tasks are also included.
When the result analysis is made, even though hybrid
algorithm is the best among many other green task scheduling
algorithms, it consumes more energy when compared to the
ESF-ES algorithm and most-efficient-server first scheme. The
ESF-ES algorithm consumes more energy than the most-
efficient-server first scheme. By comparing all the three
algorithms, the result obtained is that the most-efficient-server
first scheme algorithm is best in conserving energy in servers
of cloud data centers. In this way, task scheduling can be done
in an efficient way and the tasks are allocated to servers in
such a way that the energy consumption is greatly reduced.
REFERENCES
[1] Junwei GE, Yongsheng Y., (2013) “Research of cloud
computing task scheduling algorithm based on
improved genetic algorithm”, in: Proceedings of the
2nd International Conference on Computer Science and
Electronics Engineering.
[2] Chawla Y. and Bhonsle M., (2012) “A Study on
Scheduling Methods in Cloud Computing”,
International Journal of Emerging Trends &
Technology in Computer Science (IJETTCS) Volume
1, Issue 3, September – October.
[3] Truong V.T.D., Sato Y., Inoguchi Y., (2010)
“Performance evaluation of a green scheduling
algorithm for energy savings in cloud computing”, in:
Proc. 2010 IEEE International Symposium on Parallel
& Distributed Processing, Workshops and Phd Forum
(IPDPSW), April, pp. 1–8.
[4] Beloglazov A., Buyya R., Lee Y.C., Zomaya A., (2011)
“A Taxonomy and Survey of Energy-Efficient Data
Centers and Cloud Computing Systems”, Advances in
Computers, Volume 82, Pages 47-111.
[5] Lepakshi V.A., Dr. Prashanth C S R, (2013) “A Study
on Task Scheduling Algorithms in Cloud Computing”,
International Journal of Engineering and Innovative
Technology (IJEIT) Volume 2, Issue 11.
[6] Liu N., Dong Z., Rojas-Cessa, R., (2013) "Task
Scheduling and Server Provisioning for Energy-
Efficient Cloud-Computing Data Centers," Distributed
Computing Systems Workshops (ICDCSW), 2013
IEEE 33rd International Conference on , vol., no.,
pp.226,231, 8-11.
[7] Zhang L.M., Li K., Lo D.C., Zhang Y., (2013)
“Energy-efficient task scheduling algorithms on
heterogeneous computers with continuous and discrete
speeds”, Sustainable Computing: Informatics and
Systems 3, 109-118.
[8] http://en.wikipedia.org/wiki/Data_center.
[9] http://searchitchannel.techtarget.com/feature/Upgradin
g-to-energy-efficient-servers
[10] Zhang L.M., Li K., Zhang Y.-Q., (2010) “Green task
scheduling algorithms with speeds optimization on
heterogeneous cloud servers”, in: Proc. of 2010
IEEE/ACM International Conference on Green
Computing and Communications (Green-Com2010),
Hangzhou, December 18–20, pp. 76–80.
[11] Zhang L.M., Li K., Zhang Y.-Q., (2010) “Green task
scheduling algorithms with energy reduction on
heterogeneous computers”, in: Proc. of 2010
International Conference on Progress in Informatics
and Computing (PIC-2010), Shanghai, December 10–
12, pp. 560–563.
[12] Li K., (2008) “Performance analysis of power-aware
task scheduling algorithms on multiprocessor
computers with dynamic voltage and speed”, IEEE
Transactions on Parallel and Distributed Systems 19
(11) 1484–1497.
[13] Omara F.A., Arafa M.M., (2010) “Genetic algorithms
for task scheduling problem, Journal of Parallel and
Distributed Computing”, Volume 70, Issue 1, Pages
13–22 http://dx.doi.org/10.1016/j.jpdc.2009.09.009.
[14] Rizvandi N.B., Taheri J., Zomaya A.Y., (2011) “Some
observations on optimal frequency selection in DVFS-
based energy consumption minimization”, Journal of
Parallel and Distributed Computing 71 (8) 1154–1164.
[15] Rizvandi N.B., Taheri J., Zomaya A.Y., Lee Y.C.,
(2010) “Linear combinations of DVFS-enabled
processor frequencies to modify the energy-aware
scheduling algorithms”, in: Proc.10th IEEE/ACM
International Conference on Cluster, Cloud and Grid
Computing, pp. 388–397.
[16] Zhu D., Melhem R., Childers B.R., (2003) “Scheduling
with dynamic voltage/speed adjustment using slack
reclamation in multiprocessor real-time systems”, IEEE

__________________________________________________________________________________________
Transactions on Parallel and Distributed Systems 14 (7)
686–700.
[17] Wang L., Khan S.U., Chen D., Kolodziej J., Ranjan R.,
Xu C., Zomaya A., (2013) “Energy-aware parallel task
scheduling in a cluster”, Future Generation Computer
Systems, Volume 29, Issue 7, Pages 1661-1670.
[18] Lee Y.C., Zomaya A.Y., (2009) “On effective slack
reclamation in task scheduling for energy reduction”,
Journal of Information Processing Systems 5 (4) 175–
186.
[19] Shin Y., Chois K., Sakurait T., (2000) “Power
optimization of real-time embedded systems on
variable speed processors”, in: IEEE/ACM
International Conference on Computer Aided Design,
pp. 365–368.
[20] Ma Y., Gong B., Sugihara R., Gupta R., (2012)
“Energy-efficient deadline scheduling for
heterogeneous systems”, Journal of Parallel and
Distributed Computing, 72, 1725–1740.
[21] Kamthe A., Lee S.-Y., (2007) “Stochastic approach to
scheduling multiple divisible tasks on a heterogeneous
distributed computing system”, in: Proc. IEEE
International Parallel and Distributed Processing
Symposium, pp. 1–11.
[22] Ucar B., Aykanat C., Kaya K., Ikincib M., (2006)
“Task assignment in heterogeneous computing
systems, Journal of Parallel and Distributed
Computing”, Volume 66, Issue 1, Pages 32-46.

Energy efficient task scheduling algorithms for cloud data centers

Recommended

More Related Content

What's hot (11)

Similar to Energy efficient task scheduling algorithms for cloud data centers (20)

More from eSAT Journals (20)

Recently uploaded (20)

Energy efficient task scheduling algorithms for cloud data centers