A Meta-Analytic Review of Services Composition Methods for Cloud based Medical Applications
Babu Rajendiran1*, Jayashree Kanniappan2
1,2Rajalakshmi Engineering College, Thandalam, Chennai, India
1[email protected], 2[email protected]
ABSTRACT
In the cloud computing model user can access resources according to their requirements as a utility service. The proliferation of cloud computing leads to an increasing demand in service composition (SC) as a single service not comply with all user requirements and its critical need in practical use. Hence SC is a rapid paradigm maximizing the opportunities for reusing essential services for various medical applications. This paper addresses the various heuristic and meta-heuristic algorithms that have been utilized for the aforementioned prototype. Challenges faced during this research are identified.
Keywords: Cloud Computing, Service Composition, Heuristic and meta-heuristic algorithms, Utility service
Introduction
Cloud computing is a new standard in IT built over the vast spread of internet resources on the web and provided as a utility service to its consumers. The cloud service providers (CSPs) enable us to consume resources from an outsized IT service model for distributed network environments.
[2], [3]. There are different models of cloud services (CSs) namely Software, Platform, Infrastructure and Everything (as Service). The organizations need not to invest huge amount of money by implementing any of these service models since the computational infrastructure, maintenance and SLAs are taken care by the third party CSPs. CSs are growing rapidly because of the adoption of advanced technologies to the organizations that helps the companies to invest quality time in research and development and more purposeful tasks [18]. The highlights of cloud computing includes its dynamic nature, heterogeneity, cost effectiveness, continuous availability, openness, backup and recovery, scalability, resiliency and redundancy, and ease of integration made the CSs and CSPs selection, a key issue in this research. [20]
Cloud Consumers (CCs) habitually find it difficult to pick a CS that satisfies their functional and non-functional requirements [21]. At times a single service may not provide all desirable features that the consumers are expecting which results in the rejection of certain CSs. Consequently, it becomes more important to combine different services together and develop it as a new service called composite service that fulfills the same functional properties and QoS parameters specified by the consumer as different CSs having varying level of QoS. With the proliferation of CSs, it becomes a tedious task to search for an optimal CS for composition where specific services are chosen depending on the metric values assigned for each SOA [11]. The Composition of CSs is an NP-hard problem where inappropriate selection and QoS aware service composition leads to consumer disappointment, so proper care should be ensured while selecting individual services for composition [13], [24]. Hence composing various services according to the user requirement is a critical challenge.
This paper continues as: Section 2 elaborating the background study for SC that includes heuristic and meta-heuristic methods. Section 3 provides necessary related work, while in Section 4 various challenges and issues involved in SC are discussed. Section 5 presents the future research directions and concluded in section 6.
Background
Cloud Service Composition (CSC) utilizes heuristic and meta-heuristic algorithms to find optimal services by composing services of similar nature. This section describes the different heuristic and meta-heuristic algorithms used for CSC.
Heuristic Algorithms
Heuristic algorithms are designed for certain optimization problem that identifies an optimal solution using trial and error in a limited timeframe. Some popular heuristic algorithms are hill climbing, stochastic hill climbing, A* search and memetic algorithm which are elucidated as follows.
1. Hill Climbing
Hill climbing method generates an initial candidate solution by a random approach. Next, in all iterations the neighboring solution is differentiated with the current solution in which former differs in only one task assignment from the latter. The neighboring solution with higher utility value is found at the end of each iteration, and the iteration stops if there is no improvement in the current solution is required [19].
2. Stochastic Hill Climbing
The two components of this algorithm are candidate generator that maps one candidate solution to a set of potential descendants and an evaluation criterion that ranks each valid solution for the purpose of improving the evaluation that directs to better solutions [16].
3. A* Search
A* algorithm uses a heuristic function h(n) to determine the cost involved in reaching the service selected for composition from the current service and a function g(n) to estimate the cost from the beginning service to the current service. Therefore, the cost required for search is given by f(n)
= g(n) + h(n). The search process can be improved by selecting a good h(n) by which the solution can be achieved in short duration [22].
4. Memetic Algorithm
Memetic algorithm works on the concept of meme that means a unit of cultural evolution to represents a learning or development strategy to extract local refinement. Memetic algorithm is an extension of evolutionary algorithms to combine from different CSs with individual learning where a unique methodology is utilized to refine individual services by improving their fitness value that balances well between generality and problem specificity [23].
Meta Heuristics Algorithms
A meta-heuristic algorithm possess two main components namely intensification and diversification that makes it problem independent and suitable for various problems. The
algorithm searches for multiple solutions from the solution space and intensifies the search in the proximity of the optimal or near optimal solutions and balances between the two components as it’s significant for an operative and effectual algorithm. Examples are Genetic Algorithm, Tabu Search, Ant Colony Optimization, Cuckoo search, Gray Wolf Optimizer and Artificial Bee Colony algorithm which are described in this section.
1. Genetic Algorithm
Genetic algorithm uses the scheduling mechanism in which the tasks are allocated to resources based on the fitness value for each attribute of the selection composition process. GA works by these five phases namely Initial Population generation from the list of CSs, Fitness Function computation by comparing with neighbor, Selection of appropriate services, Crossover among available services, and Mutation of composite services based on which optimal solution is determined [25], [27].
2. Tabu Search
Tabu search is an intelligent optimization technique for solving combinatorial optimization problems that employs neighborhood method for search. Neighborhood is the set of all formations obtained as a result of one move i.e. traversing from present solution to neighboring solution. From the initial empty Tabu list, the elements are added in a circular fashion in subsequent search iterations [26].
3. Ant Colony Optimization
ACO works on the basis of behavior of ant colonies. Upon finding the food, the ant releases a special kind of chemical called Pheromone for communication by helping the other ants to sense the food. As this mechanism continues, the ants tend to find the shortest route by identifying the path that has large pheromone. Similarly service selection for composition can be determined by this extensible, inner parallelism and positive feedback mechanism [17].
4. Artificial Bee Colony
ABC utilizes the observed honey bees behavior that states that the colony of artificial bees belongs to three groups namely employed bees, onlookers, and scouts [28]. Selection of appropriate cloud service results in an optimal solution to the problem.
It operates as follows:
i. Initial Population development ii. Initial Fitness function calculation
iii. Ordering the population with the help of evaluated fitness function iv. Set the threshold
v. Fitness function calculation vi. Choosing the best bees
vii. Producing the new initial population
viii. Proceed to step 3 if the number of replications is not sufficient.
5. Cuckoo Search
To solve global optimization problems Cockoo Search is a better choice as it considers only smaller number of parameters. The algorithm guarantees a balance between exploration and exploitation by satisfying global convergence requirements that result in adaptation to ample optimization problems [4].
6. Gray Wolf Optimizer
GWO simulates the leadership qualities and hunting techniques of gray wolves in nature [5].
Different categories of grey wolves namely alpha, beta, delta, and omega are live in groups of five to twelve in form of a hierarchical society. The mathematical model for searching the prey by gray wolves is divided into (i) Trace, race, and approach (ii) Pursue, encircle, and harass until the prey stopped its movement and (iii) Confronting the prey.
Related Work
BharathBhushan et al. described a heuristic optimization procedure called QoS aware service composition in multiple cloud domains that uses PROMETHEE method to choose the finest service from related services with nominal number of cloud combinations [8]. BuyyaRajkumar et al. suggested a procedure utilizing Eagle Strategy to do optimization at two levels to protect the balance between investigation and exploitation in an effective way. In the exploration phase, service is selected similar to the way an eagle looks in real time for its prey and there will be a change in its behavior to intensive attacking during the second phase, which integrates an optimization methodology performing severe search locally [11].
The constraints forced during connectivity and steps to balance QoS parameters are assured by Fiore Ugo et al. [18] proposed an Optimal Fitness Aware Cloud Service Composition (OFACSC) using Modified Invasive Weed Optimization (MIWO). The authors discussed a novel skewness based approach that considers all possible QoS parameters and OFACSC integrated with MIWO to increase the rate of convergence and substantially down the computational complexity. Cao Jian et al. [12] presented a Cloud Service Broker (CBR) service to dynamically select the service and efficiently carry out middleware service to ease the service selection process like plug-in and play. Customer trust is accomplished by providing storage service, computing service and application service in the form of service integration.
A constraint-based approach employs techniques from the domain of product configuration and product line engineering describing a plausible way facilitating automatic aggregation of atomic services (AS) in order to realize complex user requirements by adopting feature models to model the inter-relationships and constraints among AS. The use of a structured model and automated analysis by Adigun Matthew et al. [1] enables multiple functionally equivalent AS to collaborate in service provisioning. NimaJafariNavimipour et al. [24] provided an overview of meta-heuristic algorithms used in cloud service composition under two divisions namely single and multiple cloud environments. Genetic algorithm and particle swarm optimization is used for composition of web service individually that have shorter runtime. Chella P.R et al. [13] uses Fruit Fly Optimization Algorithm (FOA) which finds an optimal composed service based on immune optimization with minimal SLA violations based on user preferences from paretoapplied on
initially selected weighted services and attains good fitness value, less error rate and shorter response time.
The ACO - Web Service Composition (WSC) algorithm was used by Chen Ling et al. [14] to effectively find a cloud combination containing minimum number of clouds and provides an optimal combination in all situations with minimum time consumption for computation. A Cloud service composition optimization technique was proposed by BuyyaRajkumar et al. [10] to help users with limited/no knowledge to deploy their services flawlessly. These users fix favorites using if-then rules to express their preferences conveniently which are then utilized by multi- objective evolutionary approaches for weighing and fuzzy inference system for specifically maintain and monitor their favorites for ranking.
Chen Ying et al. [15] combined Pareto-based techniques with Vector Ordinal Optimization (VOO) considerQoS dependency to deliver dependency-aware service composition to make the approach more concrete and common to search for optimal solutions. Bahsoon Rami et al. [6]
introduced a novel Multi-Tenant Middleware for Dynamic Service Composition integrated into a Multi-Objective Evolutionary Algorithm considering multi-tenant and QoS requirements to evaluate customized service plans for tenants as and when required.
Various researchers are using different versions of meta-heuristic algorithms inspired from biological nature by observing a group of animals to monitor its intelligence behavior and obtained better results to reach near global optimal solutions [7], [9].
Challenges in Cloud Services composition
In the cloud e-marketplace, ecosystem of AS from multiple sources has the ability to analyze numeric or text-like attributes, whereas it is difficult to cater for qualitative QoS attributes like security, user and eco-friendliness. Identifying suitable solution encoding schemes and more operational optimization algorithms to cater to the composition of Smart Grid services and increasing number of IoT is becoming more difficult.
Other challenging problem is to consider semantic information in CSC especially in a heterogeneous run time environment and should consider bio-inspired optimization methods. All interdependencies and associations among cloud services must be considered with QoS factors.
Instead of assuming all come from the same repository the researches should work to minimize the communication hindrance among clouds by considering the factor that service composition needs to be done for data from multiple service repositories.
Further Research Directions
Cloud computing provides IT enabled services to end users on a pay per use, on-request manner.
With the continuous creation of cloud computing, services are delivered with varying level of QoS. In a distributed and dynamic environment the cloud service selection and SC must try to include semantic information along with QoS attributes to find a near-optimal solution that satisfies both mandatory and QoS requirements. Multi tenancy and SLA violations also needs be considered for near optimal composition.
Conclusion
Cloud computing is characterized by its elasticity, on demand services and the unlimited provided resources. The consumer requirements are not always satisfied by providing a single service as it does not meet functional and non-functional requirements. Hence it becomes essential to combine different individual services to form a composite one to serve the requested task as any medical application is related to the other to provide better and accurate results.
Composing multiple services requires appropriate algorithms to find whether the workflows of two services are composed exactly or not. In this paper, we reviewed some existing methods for SC in cloud computing, elaborated each method and its usages.
References
[1] Adigun Matthew, DaramolaOlawande, EzenwokeAzubuike (2017), Towards a Fuzzy-oriented Framework for Service Selection in Cloud e-Marketplaces,pp. 632-637.
[2] Adnan Zuraidy, IthingMaalik, Rauf Fairuz (2018). Personal Cloud Storage using Raspberry Pi.
International Journal of Computer Applications,Vol. 180,pp. 5-8.
[3] AlnaseeriMohanaad, FraishAwatif, Hamda, Hamed, Khalil Aaisha(2019), Improving Authentication system for public cloud Computing,pp. 1-66.
[4] Altoaimy Lina, EzzatFadwa, Hassan Ahmed Syed, KurdiHeba, Youcef-Toumi Kamal (2018),MultiCuckoo: Multi-Cloud Service Composition using a Cuckoo-Inspired Algorithm for the Internet of Things Applications, IEEE Access, pp.99.
[5] Babamir S.M, Karimi M (2017),QoS-aware web service composition using Gray Wolf Optimizer,Vol.9 , no. 1, pp.9-16.
[6] Bahsoon Rami, BuyyaRajkumar, Chen Tao, Li Ke, Kumar Satish (2018), Multi-Tenant Cloud Service Composition Using Evolutionary Optimization, pp. 972-979.
[7] Bastia Abhijit, ParhiManoranjan, Patra M, PattanayakBinod (2015), Service Composition Using Efficient Multi-agents in Cloud Computing Environment, Vol. 358, pp.357-370.
[8] BharathBhushan S, Pradeep Reddy C.H (2016), A QoS Aware Cloud Service Composition Algorithm for Geo-Distributed Multi Cloud Domain. International Journal of Intelligent Engineering and Systems,Vol.9, pp. 147-156.
[9] BouguettayaAthman, Ye Zhen, Zhou Xiaofang (2012),QoS-Aware Cloud Service Composition Based on Economic Models,pp. 111-126.
[10] BuyyaRajkumar, Dastjerdi, A.V (2014), Compatibility-Aware Cloud Service Composition under Fuzzy Preferences of Users, IEEE Transactions on Cloud Computing, vol2, pp. 1-13.
[11] BuyyaRajkumar, Gangadharan G R, JatothChandrashekar, Siva Kumar Gavvala(2018), QoS-aware cloud service composition using eagle strategy, Future Generation Computer Systems,pp.90.
[12] Cao Jian, Wang Xiaogang, Xiang Yang 2015, Dynamic cloud service selection using an adaptive learning mechanism in multi-cloud computing, Journal of systems and software, vol. 100, pp. 195- 210.
[13] Chella P.R, Savarala B.B (2017), An Improved Fruit Fly Optimization Algorithm for QoSAware Cloud Service Composition, International Journal of Intelligent Engineering and Systems,Vol.
10,pp.105-114.
[14] Chen Ling, Li Bin, Yu Qiang (2015), Ant colony optimization applied to web service compositions in cloud computing, Computers & Electrical Engineering,pp. 41.
[15] Chen Ying, Huang Jiwei, Lin Chuang, ShenXuemin. (2016), Multi-Objective Service Composition with QoS Dependencies, IEEE Transactions on Cloud Computing, PP. 99.
[16] DasguptaKousik, Dutta Paramartha, MondalBrototi(2012), Load Balancing in Cloud Computing using Stochastic Hill Climbing-A Soft Computing Approach, Procedia Technology,Vol 4,pp. 783- 789.
[17] El-Sisi Ashraf, KeshkArabi, TawfeekMedhat, TorkeyFawzy A (2013), Cloud Task Scheduling Based on Ant Colony Optimization, International Arab Journal of Information Technology,Vol.
12,pp. 64-69.
[18] Fiore Ugo, Gangadharan G R, JatothChandrashekar(2018), Optimal fitness aware cloud service composition using modified invasive weed optimization, Swarm and Evolutionary Computation,pp.
44.
[19] Honiden Shinichi, Ishikawa Fuyuki, Klein Adrian (2011), Efficient Heuristic Approach with Improved Time Complexity for Qos-Aware Service Composition, Proceedings - 2011 IEEE 9th International Conference on Web Services, ICWS 2011,pp. 436-443.
[20] Jibrinoyoyo, Yusuf (2018), Adoption of Cloud Computing in Government Institutions in Nigeria.
pp. 1-105.
[21] Kanniappan J, Rajendiran B (2021) A Generic Model for Identifying QoS Parameters Interrelations in Cloud Services Selection Ontology during Runtime, Symmetry, vol.13, no. 4, pp.563.
[22] Lama Manuel, Mucientes Manuel, Rodríguez-Mier Pablo (2011), Automatic Web Service Composition with a Heuristic-Based Search Algorithm, Proceedings - 2011 IEEE 9th International Conference on Web Services, ICWS 2011,pp. 81 - 88.
[23] Ludwig Simone (2011), Memetic algorithm for web service selection, Third Workshop on Biologically Inspired Algorithms for Distributed Systems, BADS '11,pp. 1-8.
[24] NimaJafariNavimipour, Saied Asghari (2015), Review and Comparison of Meta-Heuristic Algorithms for Service Composition in Cloud Computing,Majlesi Journal of Multimedia Processing,vol4, pp. 28-34.
[25] OmaraFatma, Safwat A (2016), Genetic-Based Task Scheduling Algorithm in Cloud Computing Environment, International Journal of Advanced Computer Science and Applications, vol.7, no.4, pp. 550-556.
[26] RangaswamySirisati, Sridhar Mandapati(2018),ATabu Stochastic Diffusion Search Based Fuzzy Scheduling in Cloud, International Journal of Research in Advent Technology, Vol.6, No.5. pp.
817-823.
[27] Sasikaladevi N. (2016), SLA based cloud service composition using genetic algorithm, International Journal of Advances in Intelligent Informatics,Vol 2,pp.77.
[28] Zhang Bin, Zhang Changsheng (2014),A Hybrid Artificial Bee Colony Algorithm for the Service Selection Problem, Discrete Dynamics in Nature and Society,Pp.1-13.
