Turkish Journal of Computer and Mathematics Education Vol.12 No.3 (2021), 877-883
Research Article
An Overview of Current Multi-Attribute Techniques in Double Auction Frameworks
Muhamad Hariz Muhamad Adnan1*, Mohd Fadzil Hassan2, Izzatdin Abdul Aziz3, Okta Nurika4,
Mohd Shahid Husain5
1*Depatment of Computing, Universiti Pendidikan Sultan Idris, Tanjung Malim, Perak, Malaysia 2,3Computer and Information Sciences Deparment,
Universiti Teknologi PETRONAS, Tanjung Malim, Perak, Malaysia 4DGIT Systems, 313 La Trobe Street Melbourne, Victoria, Australia
5College of Applied Sciences, Ministry of Higher Education (MoHE), Sultanate of Oman
mhariz@fskik.upsi.edu.my1*, mfadzil_hassan@utp.edu.my2, izzatdin@utp.edu.my3, onurika@dgitsystems.com4, mshahid.ibr@cas.edu.om5
Article History: Received: 10 November 2020; Revised: 12 January 2021; Accepted: 27 January 2021;
Published online: 05 April 2021
Abstract: Double auction is becoming the preferred negotiation protocol for cloud service negotiation due to its
economic efficiency, capability in facilitating dynamic pricing, and suitability for handling a large number of customers and service providers. However, as far as this research work is concerned, there is no framework using double auction that simultaneously and comprehensively addresses the heterogeneity of cloud services in multi-attributes negotiation. Before such a framework can be designed, suitable multi-attributes negotiation techniques and its attributes should be identified. Therefore, this paper’s objective is to distinguish and provide an overview of multi-attributes techniques used in double auction negotiations. The sources are from the Scopus Database. It is found that the current multi-attributes techniques lacked in addressing the preferential dependency, selective attributes and utility optimization simultaneously in double auction frameworks. These concerns need to be addressed in order to devise a practical framework. The future direction for double auction framework in multi-attributes negotiation is suggested.
Keywords: Multi-Attribute Auctions, Double Auction, Cloud Services, Automated Negotiation, Cloud Classification
1. Highlights
Current multi-attributes techniques unable to accommodate preferential dependency, selective attributes and utility optimization simultaneously in double auction frameworks.
Algorithm based techniques may not be efficient for automatic double auction negotiations.
The utility function namely the weighted hypercubes and the matching function can be used for multi-attribute negotiations in double auction
Surprisingly, no double auction negotiation framework for cloud services was found to apply multi-attribute technique and address the multi-attribute traits namely preferential dependency, selective attributes and utilities optimization.
2. Introduction
Heterogeneous enterprise cloud services cater different varieties of needs of enterprise clients, therefore optimal pricing is not the only attribute to consider in cloud service negotiation [1]. There are other essential attributes worth considering, which are currently not being included in the double auction negotiation frameworks [2]. Double auction is becoming the preferred negotiation protocol for cloud service negotiation due to its economic efficiency, the capability in facilitating dynamic pricing and suitable for handling a large number of customers and service providers [3, 4]. However, as far as this research work is concerned, there is no framework on cloud service negotiation using double auction protocols that comprehensively address the heterogeneity of cloud services and multi-attributes negotiation both simultaneously [5].
The selection of attributes is important to support different business level objectives (BLO) [6]. These heterogeneous cloud services are also defined as variability in resource capacity, application, valuation types, capability, process technologies, and prices [7-10]. In order to accommodate heterogeneous cloud services, the double auction negotiation framework was required to optimize multiple attributes that being negotiated such as the price, virtual central processing unit (VCPU), random access memory (RAM), storage and computation time [3, 11-18]. This is important to maximize the benefits for both customer and service provider.
The multi-attribute techniques were important to tackle automatic negotiation of heterogeneous cloud services, which is complex and challenging due to its solution space which is as extensive as n-dimensional (n>1) rather than a single dimensional line [19-21]. Also, the multi-attribute techniques were important to identify the attributes’ interdependencies in the heterogeneous cloud services, which are known as a preferential dependency [19-21].
Based on the above-mentioned discrepancy, a double auction framework that is capable to tackle both heterogeneous and multi attribute cloud services simultaneously is of vital significance.Since there is no identified double auction framework that completely addresses the heterogeneity of cloud services and negotiation with multiple attributes simultaneously,an investigation of the multi-attributes techniques in double auction framework was conducted. The novelty of this paper is the overview and discussion of the multi-attribute techniques and how these techniques can be implemented in double auction negotiation framework.
The rest of the manuscript is organized as follows: Section 2 presents the current multi-attribute negotiation techniques in double auction frameworks. Section 3 discusses the findings from the study. Finally, Section 4concludes this paper.
3. Multi-attribute Negotiation Techniques in Double Auction Frameworks Algorithm Based
A good overview of existing approaches, including the most commonly used methods are given in [15, 22, 23], The techniques concentrates on pricing attributes (price and remaining resource) and non-pricing attributes (QoS, reputation, QI, and running time) [15, 22, 23]. The resource allocations of this technique are controlled by the auctioneer agents. The auctioneer agents applied the sorting algorithms or Mean-Variance Optimization (MVO) algorithm. However, this suffers from limitations due to the selections of multi-attributes for negotiation are excessive. The user must state the terms for each attribute even though some attributes may not be of significance to the user. This limitation caused the automatic negotiation to be more difficult.
On the other hand, multi-attribute negotiation technique that relied on Neural Network (NN) and hybrid algorithms may not be efficient in the double auction [4]. For instance, the Continuous Double Auction (CDA) framework that negotiates continuouslymay have problem to efficiently utilize the hybrid algorithms in real time due to its complexity. In addition, the negotiation speed can be decreased. In a practical set-up of an automatic negotiation, the negotiation speed is important [5]. It is also important that good scalability is maintained in the negotiation [6].
Further investigation was undertaken to explore multi-attribute techniques that do not rely on complex or hybrid algorithms. The utility functions and matching functions were identified. These techniques were discovered not relying on complex or hybrid algorithms. These techniques also capable to accommodate preferential dependency and dynamic attributes.
Utility Functions
The utilify function can be classified into ordinal prederences and cardinal preferences. The utility is the sum of the agent`s value for each attribute. The value of each attribute can be represented using ordinal or cardinal preferences [24]. An ordinal preference is usually represented by the ordinal graph that uses a node to represent each possible agreement and connects the nodes to the target contract preferences [24]. The main limitation of the ordinal graph is that it is unable to express the information of how much a service is preferred over another, which is vital for the broker`s agent to do service selection. Another limitation of the ordinal graph is that it scales poorly. Meanwhile, the cardinal preferences can state how much an agent prefers one service over another [5]. Therefore, four types of cardinal preferences utility functions namely additive, UCP-nets, K-additive and weighted hyper cubes were examined as shown in Table 1.
Table 1. The descriptions, advantages, and disadvantages of the four types of cardinal preferences utility
functions
Utility Function Descriptions Advantages Disadvantages
Additive [24-26]
Weighted sum of the utility for each issue.
Simple and easy to use, has been implemented widely.
Unable to capture important phenomenon of preferential dependencies.
[24, 27] each node represents agent preference for an issue.
and its preferential dependencies.
the double auction frameworks. K-additive [24] Generalization of additive utility functions where it considers the preferential dependencies as utility. Support multi-attribute and its preferential dependencies.
Not efficient for
heterogeneous cloud services as it does not consider the higher bound and lower bound of the service. Weighted hypercubes [24, 28] Represents a utility function as a collection of hypercubes. Support multi-attribute and its preferential dependencies. Can implement attribute selection. Capable of representing any imaginable utility function with a discrete domain. Support attributes boundaries as well.
The results in double auction negotiation is unknown. The technique was not widely used.
Based on Table 1, the additive utility function was unable to capture the phenomenon of preferential dependency. The K-additive functions, on the other hand, is a generalization of the additive utility function that can capture the preferential dependency. Similarly, the weighted hypercubes are capable to capture the preferential dependency of multiple attributes [24]. In addition, the weighted hypercubes consider the boundaries of the heterogeneous cloud service attributes for fitness calculation where the lower and higher boundaries of the attributes are important in cloud service negotiation. Moreover, the function can be implemented for selective attributes as well [5]. The weighted hypercubes was also able to address customer`s utilities that are non-linear[29]. It is claimed that the customer`s utilities for multi-attributes negotiation are non-linear due to the attributes constrained by one another [29].
Matching Functions
In this subsection, the matching function for multi-attribute negotiations is further investigated. The matching function is used to sort the bids and services, match them, and clear the auction [15, 22, 23, 30-32]. The matching function determines the exchange and its terms by choosing the agents, the nature of resource, dimensionality of auction, agents` utility functions, and negotiation goals [33]. It is often a part of the auctioneer`s tasks. However, most of the current matching functions are designed to optimize single attribute, such as the price [15, 22, 23]. The matching function is illustrated in Fig.1.
Figure 1. The auctioneer sorted bids in descending order and cloud services in ascending order, based on the
price
The matching functions sort bids and services based on its value in descending order in ascending order [15, 22, 23, 30]. Then, the highest bids and the lowest service price are matched to identify the winner. The final prices are set by averaging these two prices [15, 31]. In practice, the matching functions are often aimed to maximize the social welfare [30, 32]. It has also been tested to maximize the successful allocation rate, resource utilization, and buyer satisfaction [15, 32].
Some of current matching functions used certain algorithms to determine the winner. For example, the matching function in [22] uses the MVO algorithm to determine the winner and Neural Network algorithm to identify attributes preferential dependency. Moreover, the matching function in [32] applies the Greedy approximation algorithm, while the one by [23] implements the √M-approximation algorithm for sorting the bids and services. In another case, paper [31] utilizes a fitness classification algorithm to classify the complex resources in the auction.Based on the findings, the matching function can be extended to serve multi-attribute negotiation using algorithms.
4. Discussion
What is surprising is that multi-attribute negotiation techniques that relied on algorithms and hybrid algorithms may not be efficient in the double auction such as the CDA that negotiates continuously. The findings suggest that the utility function namely the weighted hypercubes and the matching function can be used for multi-attribute negotiations in double auction. To further observe the state of current double auction negotiation framework for cloud services, the findings are summarized in Table 2.
Table 2. The Summary of the Double Auction Negotiation Frameworks for Cloud Service Negotiations
Double Auction Negotiation Frameworks Multi-Attribute Technique Multi-attribute Traits Preferential Dependency Selective Attributes Utilities Optimization X. Vilajosana, D. Lázaro, A.
A. Juan, and J. M. Marquès [30]
× × × ×
Y. Lan, W. Tong, Z. Liu, and Y. Hou,
[15]
× × ×
X. Wang, X. Wang, C. L. Wang, K. Li, and M. Huang, [22]
S. Chichin, Q. B. Vo, and R. Kowalczyk, [32] × × G. Baranwal and D. P. Vidyarthi, [23] × ×
H. Wang, Z. Kang, and L. Wang,
[31]
× × ×
N. Shinde and P. S. Kiran, [34]
× × × ×
Surprisingly, no double auction negotiation framework for cloud services was found to apply multi-attribute technique and address the multi-attribute traits namely preferential dependency, selective attributes and utilities optimization. Current double auction negotiation frameworks for cloud services were mainly optimized to calculate the cloud service utility based on a single attribute, namely the price [15, 22, 23]. The findings suggest that future work to extend the current double auction negotiation framework for multi-attribute negotiation is important. This is because the cloud services are multi-attributes as shown Figure 2.
Figure 2. Multi-attributes cloud services in negotiation [35]
The findings suggest that the utility function and matching function can be used in double auction framework for multi-attribute negotiations. The utility function should accommodate multi-attribute cloud services utilities to avoid sub-optimal results [19-21]. Furthermore, the functions should address the multi-attribute traits namely the preferential dependency, selection of attributes and utilities optimization. The preferential dependency is important in multi-attribute negotiation because the preference of an attribute over other attributes can be considered as the negotiation priority or objectives in some cases [24]. Surprisingly, the current utility functions in double auction framework are unable to accommodate the preferential dependency.
Contrary to expectations, this study also found that the current utility functions in double auction frameworks did not accommodate the selection of multi-attributes in negotiation. The current framework specified that the customer must state the terms for each attribute. In real negotiation, some attributes may not be of significance to the customer. Therefore, the negotiation of only selective attributes should be allowed in the utility function [15, 22, 23]. Moreover, it was found that the current utility functions in double auction framework for cloud service negotiations does not optimize the multi-attribute utilities. On the other hand, the current matching functions in double auction negotiation frameworks was found to sort and match single-attribute cloud services [15, 22, 30].
The traits listed in Table 2 are important for a practical double auction negotiation framework for the cloud service market. The findings suggest that the weighted hypercubes utility function technique is suitable for double auction framework where it can accommodate the important traits of multi-attributes. The weighted hypercubes is also capable to address customer`s utilities that are non-linear in multi-attribute negotiation [29]. We would suggest for future work that the double auction framework can utilize the weighted hypercubes and extended matching function for multi-attribute negotiations.
5. Conclusion
This study set out to investigate the current multi-attribute techniques in double auction framework for cloud service negotiations. The study has identified that current multi-attributes techniques unable to accommodate preferential dependency, selective attributes and utility optimization simultaneously in double auction frameworks. The study has also shown that the algorithm-based techniques may not be efficient for automatic double auction negotiations. The results of this investigation show that the utility function namely the weighted hypercubes and the matching function can be used for multi-attribute negotiations in double auction.
6. Acknowledgment
The authors would like to thank the Ministry of Higher Education (MOHE), Malaysia (ERGS Grant No: 0153AB-I27, 203/PKOMP/6730002 and MyBrain15), Universiti Pendidikan Sultan Idris and Universiti Teknologi PETRONAS for supporting this study.
References
1. M. H. Adnan and M. F. Hassan, "Dynamic Pricing for Cloud Service Negotiation," in Control Theory and its Application (CNTIA) 2015, Melaka, 2016: MALTESAS.
2. M. H. Adnan, M. F. Hassan, and I. A. Aziz, "Protocols for Agent-based Autonomous Negotiations: A Review," in International Conference on Computer & Information Sciences, KLCC, Kuala Lumpur, Malaysia, 2016.
3. X. Wang, X. Wang, H. Che, K. Li, M. Huang, and C. Gao, "An intelligent economic approach for dynamic resource allocation in cloud services," IEEE Transactions on Cloud Computing, vol. 3, no. 3, pp. 275-289, 2015.
4. D. Kumar, G. Baranwal, Z. Raza, and D. P. Vidyarthi, "A systematic study of double auction mechanisms in cloud computing," Journal of Systems and Software, vol. 125, pp. 234-255, Mar 2017, doi: 10.1016/j.jss.2016.12.009.
5. M. H. Adnan, M. F. Hassan, I. A. Aziz, and N. A. Rashid, "A Survey and Future Vision of Double Auctions-Based Autonomous Cloud Service Negotiations," Cham, 2019: Springer International Publishing, in Recent Trends in Data Science and Soft Computing, pp. 488-498. 6. M. H. Adnan, M. F. Hassan, and I. A. Aziz, "Business Level Objectives of Customer for
Autonomous Cloud Service Negotiation," Advanced Science Letters, vol. 24, no. 10, pp. 7524-7528, // 2018, doi: 10.1166/asl.2018.12971.
7. G. Lee, "Resource allocation and scheduling in heterogeneous cloud environments," UNIVERSITY OF CALIFORNIA, BERKELEY, 2012.
8. H. Zhang, H. Jiang, B. Li, F. Liu, A. V. Vasilakos, and J. Liu, "A Framework for Truthful Online Auctions in Cloud Computing with Heterogeneous User Demands," IEEE Transactions on Computers, vol. 65, no. 3, pp. 805-818, 2016, doi: 10.1109/TC.2015.2435784.
9. S. Yeo and H.-H. Lee, "Using mathematical modeling in provisioning a heterogeneous cloud computing environment," Computer, vol. 44, no. 8, pp. 55-62, 2011.
10. L. T. Thai, B. Varghese, and A. D. Barker, "Algorithms for optimising heterogeneous Cloud virtual machine clusters," in 8th IEEE International Conference on Cloud Computing Technology and Science: IEEE, 2016.
11. H. Li, C. Wu, Z. Li, and F. Lau, "Virtual machine trading in a federation of clouds: Individual profit and social welfare maximization," 2013.
12. S. K. Garg, C. Vecchiola, and R. Buyya, "Mandi: a market exchange for trading utility and cloud computing services," The Journal of Supercomputing, vol. 64, no. 3, pp. 1153-1174, 2013.
13. D. Sun, G. Chang, C. Wang, Y. Xiong, and X. Wang, "Efficient Nash equilibrium based cloud resource allocation by using a continuous double auction," in Computer Design and Applications (ICCDA), 2010 International Conference on, 2010, vol. 1: IEEE, pp. V1-94-V1-99.
14. S. Shang, J. Jiang, Y. Wu, G. Yang, and W. Zheng, "A Knowledge-based Continuous Double Auction Model for Cloud Market," in 2010 Sixth International Conference on Semantics, Knowledge and Grids, 1-3 Nov. 2010 2010, pp. 129-134, doi: 10.1109/SKG.2010.22.
15. Y. Lan, W. Tong, Z. Liu, and Y. Hou, "Multi-unit continuous double auction based resource allocation method," in 2012 Third International Conference on Intelligent Control and Information Processing, 15-17 July 2012 2012, pp. 773-777, doi: 10.1109/ICICIP.2012.6391559.
16. N. Farajian and H. Ebrahim pour-komleh, "Parallel Continuous Double Auction for Service Allocation in Cloud Computing," Computer Engineering and Applications Journal, no. Vol 4, No 2 (2015), pp. 133-142, 2015.
17. X. Shi, K. Xu, J. Liu, and Y. Wang, "Continuous double auction mechanism and bidding strategies in cloud computing markets," arXiv preprint arXiv:1307.6066, 2013.
18. I. Fujiwara, K. Aida, and I. Ono, "Applying double-sided combinational auctions to resource allocation in cloud computing," in Applications and the Internet (SAINT), 2010 10th IEEE/IPSJ International Symposium on, 2010: IEEE, pp. 7-14.
19. G. Lai, C. Li, K. Sycara, and J. Giampapa, "Literature review on multi-attribute negotiations," Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, Technical Report CMU-RI-TR-04-66, 2004.
20. P. Saripalli and G. Pingali, "Madmac: Multiple attribute decision methodology for adoption of clouds," in Cloud Computing (CLOUD), 2011 IEEE International Conference on, 2011: IEEE, pp. 316-323.
21. F. Nassiri-Mofakham, M. A. Nematbakhsh, A. Baraani-Dastjerdi, N. Ghasem-Aghaee, and R. Kowalczyk, "Bidding strategy for agents in multi-attribute combinatorial double auction," Expert Systems with Applications, vol. 42, no. 6, pp. 3268-3295, 2015.
22. X. Wang, X. Wang, C. L. Wang, K. Li, and M. Huang, "Resource Allocation in Cloud Environment: A Model Based on Double Multi-attribute Auction Mechanism," in 2014 IEEE 6th International Conference on Cloud Computing Technology and Science, 15-18 Dec. 2014 2014, pp. 599-604, doi: 10.1109/CloudCom.2014.103.
23. G. Baranwal and D. P. Vidyarthi, "A fair multi-attribute combinatorial double auction model for resource allocation in cloud computing," Journal of Systems and Software, vol. 108, pp. 60-76, 2015/10/01/ 2015, doi: http://dx.doi.org/10.1016/j.jss.2015.06.025.
24. I. Marsa-Maestre, M. Klein, C. M. Jonker, and R. Aydoğan, "From problems to protocols: Towards a negotiation handbook," Decision Support Systems, vol. 60, pp. 39-54, 4// 2014, doi: http://dx.doi.org/10.1016/j.dss.2013.05.019.
25. H. Raiffa, The art and science of negotiation. Harvard University Press, 1982.
26. R. Brafman and C. J. A. m. Domshlak, "Preference handling-an introductory tutorial," vol. 30, no. 1, pp. 58-58, 2009.
27. C. Boutilier, F. Bacchus, and R. I. Brafman, "UCP-networks: A directed graphical representation of conditional utilities," in Proceedings of the Seventeenth conference on Uncertainty in artificial intelligence, 2001: Morgan Kaufmann Publishers Inc., pp. 56-64. 28. T. Ito, M. Klein, and H. Hattori, "A Multi-Issue Negotiation Protocol Among Nonlinear Utility
Agents: A Preliminary Report," in Rational, Robust, and Secure Negotiations in Multi-Agent Systems, T. Ito, H. Hattori, M. Zhang, and T. Matsuo Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2008, pp. 25-38.
29. T. Ito, M. Klein, and H. Hattori, "A multi-issue negotiation protocol among agents with nonlinear utility functions," Multiagent Grid Syst., vol. 4, no. 1, pp. 67-83, 2008.
30. X. Vilajosana, D. Lázaro, A. A. Juan, and J. M. Marquès, "A multi-lane double auction for economic-based service management in the cloud," in Intelligent Networking, Collaborative Systems and Applications: Springer, 2010, pp. 117-148.
31. H. Wang, Z. Kang, and L. Wang, "Performance-Aware Cloud Resource Allocation via Fitness-Enabled Auction," IEEE Transactions on Parallel and Distributed Systems, vol. 27, no. 4, pp. 1160-1173, 2016, doi: 10.1109/TPDS.2015.2426188.
32. S. Chichin, Q. B. Vo, and R. Kowalczyk, "Towards Efficient Greedy Allocation Schemes for Double-Sided Cloud Markets," in 2015 IEEE International Conference on Services Computing, June 27 2015-July 2 2015 2015, pp. 194-201, doi: 10.1109/SCC.2015.35.
[Online]. Available:
https://ieeexplore.ieee.org/ielx7/7194374/7207317/07207353.pdf?tp=&arnumber=7207353&is number=7207317
33. P. R. Wurman, M. P. Wellman, and W. E. Walsh, "A Parametrization of the Auction Design Space," Games and Economic Behavior, vol. 35, no. 1, pp. 304-338, 2001/04/01/ 2001, doi: https://doi.org/10.1006/game.2000.0828.
34. N. Shinde and P. S. Kiran, "A survey of Cloud Auction mechanisms & decision making in Cloud Market to achieve highest resource & cost efficiency," in 2016 International Conference on Automatic Control and Dynamic Optimization Techniques (ICACDOT), 9-10 Sept. 2016 2016, pp. 1158-1162, doi: 10.1109/ICACDOT.2016.7877768. [Online]. Available: https://ieeexplore.ieee.org/ielx7/7871261/7877540/07877768.pdf?tp=&arnumber=7877768&is number=7877540
35. M. H. M. Adnan, "Double Auction-based Negotiation Framework for Heterogeneous and Multi-Attributes Cloud Services," PhD in Information Technology, Universiti Teknologi PETRONAS, 2019.
