Task level energy and performance assurance workload scheduling model in distributed computing environment

International Journal of Reconfigurable and Embedded Systems (IJRES)
Vol. 13, No. 1, March 2024, pp. 210~216
ISSN: 2089-4864, DOI: 10.11591/ijres.v13.i1.pp210-216  210
Journal homepage: http://ijres.iaescore.com
Task level energy and performance assurance workload
scheduling model in distributed computing environment
Jagadevi Bakka1
, Sanjeev C. Lingareddy2
1
Department of Computer Science and Engineering, East Point College of Engineering and Technology, Bangalore, India
2
Department of Computer Science and Engineering, Sri Venkateshwara College of Engineering, Bangalore, India
Article Info ABSTRACT
Article history:
Received Jul 22, 2022
Revised Jun 30, 2023
Accepted Aug 5, 2023
Scientific workload execution on a distributed computing platform such as a
cloud environment is time-consuming and expensive. The scientific workload
has task dependencies with different service level agreement (SLA)
prerequisites at different levels. Existing workload scheduling (WS) designs
are not efficient in assuring SLA at the task level. Alongside, induces higher
costs as the majority of scheduling mechanisms reduce either time or energy.
In reducing, cost both energy and makespan must be optimized together for
allocating resources. No prior work has considered optimizing energy and
processing time together in meeting task level SLA requirements. This paper
presents task level energy and performance assurance-workload scheduling
(TLEPA-WS) algorithm for the distributed computing environment. The
TLEPA-WS guarantees energy minimization with the performance
requirement of the parallel application under a distributed computational
environment. Experiment results show a significant reduction in using energy
and makespan; thereby reducing the cost of workload execution in comparison
with various standard workload execution models.
Keywords:
Cloud computing
Direct acyclic graph
MapReduce
Service level agreement
Task level energy and
performance assurance
This is an open access article under the CC BY-SA license.
Corresponding Author:
Jagadevi Bakka
Department of Computer Science and Engineering, East Point College of Engineering and Technology
Jnana Prabha, East Point Campus, Virgo Nagar Post, Avalahalli, Bengaluru, Karnataka, India
Email: jagdevi198@rediffmail.com
1. INTRODUCTION
Recently, the big data framework is being emphasized in wide-range of complex scientific
applications. For example, scientific complex workloads such as Montage (Figure 1) which are widely used
for the scientific and business investigation [1]. The significant growth has led to some serious challenges such
as low-latency and cost effectiveness for storing, communication, and processing [2]. In addressing such
problem, cloud computing such as Amazon EC2, Pegasus, and MapReduce [3], [4] are being used. Cloud
platforms provide high quality storage and computing resources like networks, services, and applications. for
the execution of the scientific complex workloads [5]. The scientific workload generally represented as a
directed acyclic graph (DAG) where there exist dependencies among task as shown in Figure 1. This makes
scheduling of cloud resource for workload execution a challenges task [6].
Recently, the researchers are using the cloud services to schedule the workload [7]–[10]. The
Figure 2 depicts a simple architecture for the workload scheduling (WS) in a cloud environment. However, to
design an efficient workload-scheduling model by reviewing the existing models presents various challenges
like executing larger and more complex scientific workloads, which requires more execution time and increases
the energy usage for the execution. It becomes more challenging when the tasks have to be executed in a given
deadline. Moreover, the scheduling of workload is deemed an NP-hard problem [11]–[14]. In fact, optimizing
both time and cost is a very challenging task in WS [15]. For an instance, if a scheduling model seeks to reduce

Int J Reconfigurable & Embedded Syst ISSN: 2089-4864 
Task level energy and performance assurance workload scheduling model in distributed … (Jagadevi Bakka)
211
the energy usage, the time required to complete a given task increases. This is because both the energy and
time are linked. Since many existing models do not consider task level service level agreement (SLA) into
account while creating schedules, the energy and makespan optimization problem persists [16]. To address the
above problem, this paper presents a task level energy and performance assurance-workload scheduling
(TLEPA-WS) technique. The TLEPA-WS assures SLA at task level [17], [18]. Research contribution for this
research is mentioned as: i) the TLEPA-WS is efficient in minimizing energy with performance assurance of
workload task level SLA prerequisite. No prior work has considered such scheduling mechanism and ii) the
TLEPA-WS model reduces energy consumption, makespan, and computational cost in comparison with
energy-min scheduling.
The paper organization: in section 2, study various existing WS design for distributed computational
environment. In section 3, the TLEPA workload-scheduling model is explained. The simulation study is given
in section 4. In last section, the research is concluded and future research enhancement is given.
Figure 1. Sample representation of Montage workload [2]
Figure 2. The standard resource-scheduling framework

 ISSN: 2089-4864
Int J Reconfigurable & Embedded Syst, Vol. 13, No. 1, March 2024: 210-216
212
2. LITERATURE SURVEY
This section studies various workload-scheduling model to leveraging distributed computing platform
such as cloud environment. Table 1 comprises of several survey regarding this research. Here the advantage as
well as limitation is mentioned of all the survey. In addressing aforementioned limitation in next section,
presents task level performance and energy assurance WS technique for cloud environment.
Table 1. Survey table
Method Advantage and limitation Metrics and workload used
Hu et al. [19] proposed an algorithm, which
reduces the scheduling length by providing
proper reliability and also another
algorithm for the processor has been
proposed which reduces the execution time
using the dynamic voltage and frequency
scaling (DVFS) method to minimize the
energy consumption.
This model has performed better in terms of
energy and computation time. Furthermore, this
model has also addressed the issue of reliability
during the execution of the tasks. However, the
given model consumes more cost for the
execution of the task of the workflow, hence is
not cost efficient.
Energy consumption and
computation time are metric used.
Montage, cybershake, laser
interferometer gravitational wave
observatory (LIGO) and some
synthetic workflows are used.
Pham and Fahringer [20] proposed a multi-
objective workflow scheduling method
which provides a trade-off solution to
reduce the cost and increase the makespan.
This method performs better in terms of
makespan and cost. However, this model has
not addressed the issue of energy consumption
during the execution of the workflow in the
clouds.
Makespan, cost, deadline, and
budget are metric used.
cybershake, epigenomics, inspiral,
Montage and sipht are workload
used.
Zhou et al. [21] used two method deadline-
constrained cost optimization for hybrid
clouds (DCOH) and multi-objective
optimization for hybrid clouds (MOH)
which optimizes the makespan and reduces
the cost.
This method has reduced the execution cost by
100% when compared with the deadline
constraint models. However, in order to reduce
the cost this model has consumed more energy
and this model is good for hybrid clouds, not
good for edge cloud platforms.
Energy, cost, and makespan are
metric used. Cybershake,
epigenomics, inspiral, Montage,
and sipht are workload used.
Wang et al. [22] presented an algorithm,
deadline and budget constrained
heterogenous scheduling (DMW-HDBS) to
select the task, check the priority of the task
so that the resources should be allocated to
the most priority task first and then to the
least priority tasks.
This model has showed good performance and
has increased the success rate for the execution
of the task in the multi-workflow scheduling
platforms using their method. However, this
model has not addressed the issue of energy
consumption by each of the task during the
execution of the workflow.
Cost, deadline, and makespan are
metric used. Lille, Sophia, and
Rennes are workflows used.
Tang [23] proposed a strategy to provide a
cost-efficient and reliability method for the
multi-cloud network. In this model, they
have developed a fault-tolerant workflow-
scheduling framework, which increase the
reliability and reduces the cost during the
execution of the task.
The results show that the model has
outperformed when compared with the existing
methods in terms of reliability and cost.
However, the model has failed to address the
issue of energy consumption in their model.
Execution time, cost and
reliability are metrics used. LIGO
and epigenomics are workload
used.
Malik et al. [24] proposed an algorithm
based on the classification of the task and
the threshold in order to provide better
resources for the execution of the task.
This model has used particle swarm
optimization (PSO) to increase the performance
of the model and reduce the energy
consumption, makespan and load balancing.
However, this model consume more time to
execute as it has to run the algorithm first and
then run the PSO algorithm to get its result,
hence, is not time-efficient method.
Makespan, energy consumption,
load balance are metric used.
Cybershake, epigenomics, LIGO,
Montage, and sipht are workflows
used.
3. ENERGY EFFICIENT MULTI-SENSORY TARGET TRACKING METHOD FOR WIRELESS
SENSOR NETWORK
This section present TLEPA-WS algorithm in cloud computing environment as shown in Figure 3.
The TLEPA provide ideal performance considering varying SLA condition at task level in parallel
computational framework. In this section, we design and task-level SLA-based WS algorithm for monitoring
the process and solves the various attributes and constraints; in here various constraint is considered such as
priority, task dependency, parallel computation and various attribute such as energy and makespan is
considered for minimization.
3.1. Algorithm
This section presents the TLEPA algorithm for execution of complex workload in distribute
computing platform in Algorithm 1. In this section, we design and task-level SLA-based WS algorithm for
monitoring the process and solves the various attributes and constraints; in here various constraint is considered
such as priority, task dependency, parallel computation and various attribute such as energy and makespan is
considered for minimization.

213
Figure 3. The workload-scheduling framework
Algorithm 1. The algorithm of adaptive workload scheduling assuring SLA at task level
Step 1. Design a task and power model.
Step 2. Compute the given lower performance constraint limits.
Step 3. Mechanism for adaptive task scheduling in which tasks that have not been scheduled are tested to see
if the necessary resources are available for their execution.
Step 4. Design the tasks in DAG. The number of levels in the DAG is w, where w is the number of levels with
varying energy and makespan SLA constraint in the DAG.
Step 5. Non-urgent tasks fall under the category of level 1.
Step 6. TLEPA organizes tasks into levels 1, 2, 3, 4, etc., and keeps track of all of them.
Step 7. Furthermore, unless the task in level m is finished, m+1 level cannot be executed. The monitoring
process is the same for each level m.
Step 8. As tasks are independent and have different levels as a result, they are scheduled using an adaptive task
scheduling technique.
3.2. Service level agreement optimization
In this given section, we calculate the lower bound. Assume a parameter X which represents the
amount of work done for a given parallel task o, this can be represented using as (1).
X = x1 + x2 + ⋯ + xo = ℙ1 s1 + ℙ1 s2 + ⋯ + ℙoso (1)
The minimal energy and optimal length required for an optimal scheduling is represented using F′
and
U′
respectively. By considering, all these parameters, the lower bound can be evaluated which help to reduce
the makespan in a multi-level workload-scheduling model. In (2) represent that the lower bound reduces the
makespan and the given (3) represents the lower bound to reduce the consumption of energy.
U′
≥ (
n
F
̃
(
X
n
)
b
)
1/(b−1)
(2)
F′
≥ n (
X
n
)
b
(U
̃b−1
)
−1
(3)
Both the (2) and (3) can be used for any dependent, independent and parallel task.
Method first considers uc as the time required for the execution of task Okd. In this section the tasks
which are not scheduled are tested and monitored just to understand whether the task is free for the execution
or not, i.e., it means to check whether the resources required for the execution of the task are available or not.
Hence, this results in including a greater number of tasks than the previous methods [19], [25]. Furthermore,
the scheduling of the task helps to improve the utilization of the processor. The scheduling of the task can be
represented using the equation, i.e., when the task O=1 then:

 ISSN: 2089-4864
214
Q∗(O, N) = ∑ nsn
N
n=1 (4)
suppose the task O is more than 1 then it is represented using the (5).
Q∗(O, N) = ∑ sn(n + Q∗
(O − 1, N − n))
N
n=1 + (∑ sn
n>N )Q∗
(O − 1, N) (5)
If the size of the first task is large, then the resources, which are available, are used and all the other
tasks are terminated. After this all the other remaining task 0 − 1 are scheduled. Further average scheduling
after the above situations is Q∗(O − 1, N). The proposed model resource utilization efficiency is studied
through experiment analysis in next section.
4. EXPERIMENT RESULT AND ANALYSIS
This section studies the performance in terms of makespan, energy efficiency and cost efficiency
achieved using proposed Amazon web services-transport layer security (AWS-TLS) over existing energy-
minimized scheduling of real-time parallel workflows (EMS-RTPW) [19] model. The AWS-TLS and EMS is
implemented using Java programing language using cloudsim [2], [3]. The memory intensive nature inspiral
is used for validating workload-scheduling models. Makespan, energy, and cost efficiency are metrics used for
validating workload-scheduling models.
4.1. Makespan performance
Here the makespan efficiency of TLEPA and EMS is measured by varying the inspiral workload task
size from 30 (small), 50 (large), and 100 (large) as shown in Figure 4. An average makespan reduction of
80.33% is achieved using TLEPA-WS over EMS-WS. All the observation can be obtained from the graph.
Figure 4. Makespan efficiency with different inspiral workload size
4.2. Energy efficiency
Here the energy efficiency of TLEPA and EMS is measured by varying the inspiral workload task
size from 30 (small), 50 (large), and 100 (large) as shown in Figure 5. An average energy consumption
reduction of 44.96% is achieved using TLEPA-WS over EMS-WS. All the observation can be obtained from
the graph.
Figure 5. Energy efficiency with different inspiral workload size

215
4.3. Cost efficiency
Here the cost efficiency of TLEPA and EMS is measured by varying the inspiral workload task size
from 30 (small), 50 (large), and 100 (large) as shown in Figure 6. An average cost reduction of 80.1% is
achieved using TLEPA-WS over EMS-WS. All the observation can be obtained from the graph.
Figure 6. Cost efficiency with different inspiral workload size
5. CONCLUSION
This paper presented an efficient WS that assures task level SLA. No prior work has considered
scheduling workload considering task-level SLA. The model can reduce energy and maintain high-level of
performance; thereby significantly reducing overall execution cost. Experiment outcome shows the TLEPA-
WS is very efficient in terms of energy efficiency i.e., an improvement of 44.96% is experienced by TLEPA-
WS over EMS-WS. The makespan for execution of workload is reduced by 83.33% by TLEPA-WS over EMS-
WS. Similarly, cost for execution of workload is reduced by 80.1% by TLEPA-WS over EMS-WS. The future
work would test the model considering different resource intensive workload such as central processing unit
(CPU), I/O and memory. Alongside, improve scheduling considering much larger complex task.
REFERENCES
[1] L. Wu, R. Ding, Z. Jia, and X. Li, “Cost-effective resource provisioning for real-time workflow in cloud,” Complexity, vol. 2020,
pp. 1–15, Mar. 2020, doi: 10.1155/2020/1467274.
[2] L. Chen, X. Li, Y. Guo, and R. Ruiz, “Hybrid resource provisioning for cloud workflows with malleable and rigid tasks,” IEEE
Transactions on Cloud Computing, vol. 9, no. 3, pp. 1089–1102, Jul. 2021, doi: 10.1109/TCC.2019.2894836.
[3] J. Wang, X. Li, R. Ruiz, J. Yang, and D. Chu, “Energy utilization task scheduling for mapreduce in heterogeneous clusters,” IEEE
Transactions on Services Computing, vol. 15, no. 2, pp. 931–944, Mar. 2022, doi: 10.1109/TSC.2020.2966697.
[4] Z. Wen, R. Qasha, Z. Li, R. Ranjan, P. Watson, and A. Romanovsky, “Dynamically partitioning workflow over federated clouds
for optimising the monetary cost and handling run-time failures,” IEEE Transactions on Cloud Computing, vol. 8, no. 4, pp. 1093–
1107, Oct. 2020, doi: 10.1109/TCC.2016.2603477.
[5] A. A. Nasr, N. A. El-Bahnasawy, G. Attiya, and A. El-Sayed, “Cost-effective algorithm for workflow scheduling in cloud computing
under deadline constraint,” Arabian Journal for Science and Engineering, vol. 44, no. 4, pp. 3765–3780, 2019, doi: 10.1007/s13369-
018-3664-6.
[6] X. Zhou, G. Zhang, J. Sun, J. Zhou, T. Wei, and S. Hu, “Minimizing cost and makespan for workflow scheduling in cloud using
fuzzy dominance sort based HEFT,” Future Generation Computer Systems, vol. 93, pp. 278–289, 2019, doi:
10.1016/j.future.2018.10.046.
[7] G. Ismayilov and H. R. Topcuoglu, “Neural network based multi-objective evolutionary algorithm for dynamic workflow
scheduling in cloud computing,” Future Generation Computer Systems, vol. 102, pp. 307–322, 2020, doi:
10.1016/j.future.2019.08.012.
[8] A. M. Manasrah and H. B. Ali, “Workflow Scheduling using hybrid GA-PSO algorithm in cloud computing,” Wireless
Communications and Mobile Computing, vol. 2018, pp. 1–16, 2018, doi: 10.1155/2018/1934784.
[9] S. Yassir, Z. Mostapha, and T. Claude, “Workflow scheduling issues and techniques in cloud computing: a systematic literature
review,” CloudTech 2017: Cloud Computing and Big Data: Technologies, Applications and Security, 2019, pp. 241–263. doi:
10.1007/978-3-319-97719-5_16.
[10] Y. Qin, H. Wang, S. Yi, X. Li, and L. Zhai, “An energy-aware scheduling algorithm for budget-constrained scientific workflows
based on multi-objective reinforcement learning,” Journal of Supercomputing, vol. 76, no. 1, pp. 455–480, 2020, doi:
10.1007/s11227-019-03033-y.
[11] J. K. Konjaang and L. Xu, “Cost optimised heuristic algorithm (COHA) for scientific workflow scheduling in IaaS cloud
environment,” Proceedings - 2020 IEEE 6th Intl Conference on Big Data Security on Cloud, BigDataSecurity 2020, 2020 IEEE
Intl Conference on High Performance and Smart Computing, HPSC 2020 and 2020 IEEE Intl Conference on Intelligent Data and
Security, IDS 2020, 2020, pp. 162–168, doi: 10.1109/BigDataSecurity-HPSC-IDS49724.2020.00038.
[12] Z. Li, L. Zhang, Y. Cai, and H. Ochiai, “Sensor selection for maneuvering target tracking in wireless sensor networks with
uncertainty,” IEEE Sensors Journal, vol. 22, no. 15, pp. 15071–15081, Aug. 2022, doi: 10.1109/JSEN.2021.3136546.

 ISSN: 2089-4864
216
[13] F. Liu, C. Jiang, and W. Xiao, “Multistep prediction-based adaptive dynamic programming sensor scheduling approach for
collaborative target tracking in energy harvesting wireless sensor networks,” IEEE Transactions on Automation Science and
Engineering, vol. 18, no. 2, pp. 693–704, Apr. 2021, doi: 10.1109/TASE.2020.3019567.
[14] S. R. Jondhale and R. S. Deshpande, “Kalman filtering framework-based real time target tracking in wireless sensor networks using
generalized regression neural networks,” IEEE Sensors Journal, vol. 19, no. 1, pp. 224–233, Jan. 2019, doi:
10.1109/JSEN.2018.2873357.
[15] Y. Zhang and L. Gao, “Sensor-networked underwater target tracking based on grubbs criterion and improved particle filter
algorithm,” IEEE Access, vol. 7, pp. 142894–142906, 2019, doi: 10.1109/ACCESS.2019.2943916.
[16] Y. Wang, H. Liu, J. Zhou, and Y. Wan, “Distributed target tracking in sensor networks by consistency algorithm and semantic
moving computing of internet of things,” IEEE Access, vol. 10, pp. 51634–51642, 2022, doi: 10.1109/ACCESS.2022.3170477.
[17] N. D. Zade, S. N. Deshpande, and D. Sita, “Investigative analysis of suboptimal filter for state estimation in object tracking wireless
sensor network,” IEEE Sensors Letters, vol. 4, no. 10, pp. 1–4, Oct. 2020, doi: 10.1109/LSENS.2020.3026263.
[18] B. N. B. Reddy, B. Pardhasaradhi, G. Srinath, and P. Srihari, “Distributed fusion of optimally quantized local tracker estimates for
underwater wireless sensor networks,” IEEE Access, vol. 10, pp. 38982–38998, 2022, doi: 10.1109/ACCESS.2022.3164515.
[19] B. Hu, Z. Cao, and M. Zhou, “Energy-minimized Scheduling of real-time parallel workflows on heterogeneous distributed
computing systems,” IEEE Transactions on Services Computing, vol. 15, no. 5, pp. 2766–2779, Sep. 2022, doi:
10.1109/TSC.2021.3054754.
[20] T.-P. Pham and T. Fahringer, “Evolutionary multi-objective workflow scheduling for volatile resources in the cloud,” IEEE
Transactions on Cloud Computing, vol. 10, no. 3, pp. 1780–1791, Jul. 2022, doi: 10.1109/TCC.2020.2993250.
[21] J. Zhou, T. Wang, P. Cong, P. Lu, T. Wei, and M. Chen, “Cost and makespan-aware workflow scheduling in hybrid clouds,” Journal
of Systems Architecture, vol. 100, pp. 1–12, Nov. 2019, doi: 10.1016/j.sysarc.2019.08.004.
[22] G. Wang, Y. Wang, M. S. Obaidat, C. Lin, and H. Guo, “Dynamic multiworkflow deadline and budget constrained scheduling in
heterogeneous distributed systems,” IEEE Systems Journal, vol. 15, no. 4, pp. 4939–4949, Dec. 2021, doi:
10.1109/JSYST.2021.3087527.
[23] X. Tang, “Reliability-aware cost-efficient scientific workflows scheduling strategy on multi-cloud systems,” IEEE Transactions on
Cloud Computing, vol. 10, no. 4, pp. 2909–2919, Oct. 2022, doi: 10.1109/TCC.2021.3057422.
[24] N. Malik, M. Sardaraz, M. Tahir, B. Shah, G. Ali, and F. Moreira, “Energy-efficient load balancing algorithm for workflow
scheduling in cloud data centers using queuing and thresholds,” Applied Sciences, vol. 11, no. 13, pp. 1–17, Jun. 2021, doi:
10.3390/app11135849.
[25] M. Barika, S. Garg, A. Chan, and R. N. Calheiros, “Scheduling algorithms for efficient execution of stream workflow applications
in multicloud environments,” IEEE Transactions on Services Computing, vol. 15, no. 2, pp. 860–875, Mar. 2022, doi:
10.1109/TSC.2019.2963382.
BIOGRAPHIES OF AUTHORS
Jagadevi Bakka working as assistant professor in Department of Computer Science
and Engineering, East Point College of Engineering and Technology, Bangalore. Her teaching
experience is 18 years and her area of interest is cloud computing, big data analytics, data mining,
software engineering, and data structures. She can be contacted at email:
jagdevi198@rediffmail.com or jagadevi.bakka@gmail.com.
Dr. Sanjeev C. Lingareddy received his Ph.D. in the year of 2012 from JNTU,
Hyderabad and currently working as professor and Head for the Department of Computer
Science and Engineering at Sri Venkateshwara College of Engineering, Bengaluru. He has 24
years of rich experience in the Academics and 7 years of research experience. He has published
more than 25 research articles in International Journals. He is a Member of Indian Society for
Technical Education (MISTE) and an active member in many technical events. His research area
includes Wireless sensor network, wireless security, cloud computing and cognitive network.
He can be contacted at this email: sclingareddy@gmail.com.

Task level energy and performance assurance workload scheduling model in distributed computing environment

More Related Content

Similar to Task level energy and performance assurance workload scheduling model in distributed computing environment (20)

More from International Journal of Reconfigurable and Embedded Systems (20)

Recently uploaded (20)

Task level energy and performance assurance workload scheduling model in distributed computing environment