Efficient task scheduling strategies using symbiotic organisms search algorithm for cloud computing environment
In recent times, the cloud computing model is gaining tremendous migration of users in both private and government organisations. Users are charged based on their resources usage as well as Quality of Service (QoS) desired due to its pay-as-you-go feature. As such, task scheduling approaches play...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/103977/1/FSKTM%202022%203%20IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In recent times, the cloud computing model is gaining tremendous migration of users in
both private and government organisations. Users are charged based on their resources
usage as well as Quality of Service (QoS) desired due to its pay-as-you-go feature. As
such, task scheduling approaches play a vital role in specifying and ensuring an adequate
set of resources to execute the users’ applications (i.e., tasks). Hence, schedule decisions
using task scheduling approaches are according to the users’ outlined QoS requirements.
Task scheduling in the cloud environment is an NP-Complete problem, because of its
dynamism and huge search space given by problem instances that are large scale.
Although many approaches (heuristics and metaheuristics) were proposed, they incur
high computational complexity as well as not that efficient in terms of obtaining global
optimum solutions. Recently, a nature-inspired metaheuristic known as Symbiotic
Organisms Search (SOS) optimisation algorithm was proposed. It imitates the survival
relationships (mutualism, commensalism and parasitism) between organisms in an
ecosystem.
The SOS and its variants Discrete Symbiotic Organisms Search (DSOS) algorithm have
been used to solve different optimisation problems including tasks scheduling in cloud
computing environment where results obtained are promising in comparison with stateof-
the-art metaheuristic algorithms. However, the efficiency of the clouds drops as the
size of the search space gets larger, like in the case of most metaheuristic optimisation
algorithms. Furthermore, existing tasks scheduling solutions suffer from local optima
entrapment due to inadequate diversification of their local search space, a high degree of
imbalance because of their static control parameters cannot maintain a balance between
local and global search space and scalability issue as a result of static benefit factors. In this study, an enhanced Discrete Symbiotic Organisms Search (eDSOS), a Cuckoobased
Symbiotic Organisms Search (CDSOS), and an Adaptive Symbiotic Organisms
Search (ADSOS) approaches are proposed to address the issues of local optima
entrapment, load balancing as well as scalability due to large scale task scheduling
optimisation problem in IaaS cloud computing environment. To solve the issue of local
optima entrapment, the concept of diversifying the local search space by enhancing SOS
named eDSOS approach to avoid entrapment in local optima for global convergence.
Then CDSOS approach further improves the SOS algorithm by hybridising it with
Cuckoo search’s levy flight to minimise the degree of imbalance between the local search
space and the global search thereby improving the tasks to VM mapping. Finally, to solve
the issue of scalability ADSOS approach adaptively turn SOS benefit factors to make
SOS more efficient for solving large scale task scheduling problems as well as faster
convergence speed. To assess the effectiveness of the proposed approaches (eDSOS,
CDSOS, and ADSOS) CloudSim simulator was used, using synthesised workloads
(normal, left-half, right-half and uniform distributions). Moreover, a comparison of the
proposed approaches was done with DSOS, SASOS and OTB-CSO, respectively.
The proposed approaches obtained considerable improvement in terms of the following
metrics: makespan, response time, degree of imbalance, execution cost and execution
time while meeting the desired QoS requirements. Furthermore, the simulation results
showed that the eDSOS task scheduling approach outperformed the benchmark
algorithm, produced better makespan time performance of 15.93%, 16.22%, 19.69% and
14.54% whereas the benchmark algorithm produced 18.04%, 19.64%, 16.08% and
14.72%, when implemented on the same dataset. Also, the results of the simulations on
normal, left-half, right-half and uniform datasets showed the proposed scheduling
approach obtained a better performance on degree of imbalance over the benchmarked
OTB-CSO algorithm. Moreover, the results of the simulations also showed the ADSOS
produced Performance Improvement Rate (PIR%) of 39.45%, 35.08% and 23.91%,
21.36% compared to the benchmarked algorithm in term of execution cost and execution
time respectively. Hence, these gives a superior middle way between the execution of
cost and time which makes it be dependable for its implementation in a real cloud
computing environment. Consequently, the approaches proposed have abilities to better
the QoS delivery. The research made some recommendations such as to implement the
proposed SOS-based task scheduling approaches using NASA Ames iPSC/860 and
HPC2N workloads as well a real cloud environment to validate their performances. |
|---|