Quality of Service in Service Oriented Architecture Applications - Research Proposal Example

?Quality of Service (QoS) in Service Oriented Architecture (SOA) Applications { } { address } { email address } – This document gives an account of mathematical models for Quality of Service (QoS) for Service Oriented Architectures (SoAs) in terms of reliability and performance. Currently there are multiple ways of defining both reliability and performance for QoS in SoAs. This paper will look into reliability and performance models for SOA developed for multi-tier inventory control system using simulations of the mathematical models. The current levels of available reliability and performance will be assessed for their application to real life business applications. Keywords – Quality of Service, Service Oriented Architecture, QoS, SOA, reliability, performance, mathematical model I. Introduction Service oriented architecture (SOA) environments have gained popularity in recent years given their inherent flexibility and reusability. The typical SOA is structured around methods that promote software creation as interoperable services [1]. The various services used to construct a SOA are well-defined business functions. These business functions are construed as various software building blocks that can be used repeatedly to promote decreased development time and debugging effort [2]. Currently the SOA model is finding great appeal for web application development as well as grid computing [3]. The current speed of SOA architecture development indicates that SOA applications will dominate the future avenues of development. II. Quality of Service (QoS) and Service Oriented Architecture (SOA) The most important method of describing a SOA’s characteristics in general is the quality of service (QoS) it offers [4]. The current development of the QoS concept sees divergence and some commonality depending on the various fields that employ SOAs. Most web based application SOAs require real time transfer rates such as for fiscal information transfer or multimedia applications. In contrast, other areas such as mobile grid computing see SOA QoS as a set of non-functional outputs that guarantee reliability [5]. The existing disparity between QoS definitions for SOA implementations in networking, web applications and grid computing mean that QoS needs to be defined in a more comprehensive yet unambiguous fashion. Most previous definitions of QoS promote the idea that only the best possible service levels can be classified as QoS [6]. However, more modern definitions of QoS contend that any level of services that meet user requirements can be classified as QoS for that particular application [7]. QoS can also be compared to Service Level Agreements (SLAs) where the user and the service provider agree upon acceptable levels of performance parameters such as functionality, costs etc. QoS for SOAs can be classified in the form of “request, specification, provision and negotiation” of some particular network features [7]. A typical SOA application may require QoS for the latency, the error rate, bandwidth, availability as well as the network security [8]. However, these characteristics alone should not be seen as describing the entire umbrella for QoS for SOAs. Instead, these characteristics form some of the characteristics of QoS for SOAs. Moreover, the QoS agreements would tend to vary from application to application in SOA depending on user requirements, functional expectations and non-functional expectations of the user [9]. For example, users accessing multimedia content through SOA applications would emphasize more on bandwidth than users dealing with defense applications through SOA who would emphasize more on reliability and security [10]. The classification method provided above for QoS in SOA applications is limited in scope. The QoS may be broken down into smaller metrics for further QoS classification. For example, jitter in network connection could be treated through consecutive difference in jitter or through the average jitter. The results from these metrics would tend to differ from one application to the other and further research is the only real method of specifying which QoS metric would fit which application [10]. Another way of looking at QoS would be to see QoS applying to the boundaries of the overall SOA application rather than to specific parts of the application [11]. This would tend to produce a QoS definition easier to generate and process but may not provide the insight required to deal with the application [12]. Hence, it could be argued that wholesome methods of looking at QoS would be unsuitable for SOA applications since they depend on a number of characteristics that cannot be summed up whole or be represented by a single metric. QoS research by certain groups also suggests that certain core QoS characteristics can be produced which apply equally to all kinds of SOA applications [13]. Current estimates of these core characteristics include (but are not limited to) dependability, accuracy, speed and failure semantics. Dependability is classified further into reliability, safety, availability and security for SOA purposes [14]. III. Mathematical Models for Reliability and Performance in SOAs This paper aims to look at two distinct parameters for QoS in SOA applications namely reliability and performance. Although reliability could be defined simply such as through the loss of network traffic at certain thresholds but performance can be defined in a number of different manners [15]. This paper does not attempt to encompass all the possible ways of looking at performance for SOAs but provides one perspective of looking at performance for SOAs. The mathematical models for SOA implementation developed by Zeng, Huang and Fan (2009) for assessing reliability and performance will be used for the purposes of this paper. As SOAs develop, the growth of service-oriented enterprises (SOEs) and service oriented enterprise networks (SOENs) are becoming common for enterprise integration (EI). The business model proposed by Zeng, Huang and Fan (2009) relies on four distinct layers that can be used to simulate business processes. These layers include the enterprise layer, the process / service layer, the data layer and the application layer. In addition, business processes that are dealt with using SOA implementation are classified as depending on networked services (NS). Each networked service corresponds to some form of business process that adds value to the overall business. The NS are further divided into three levels depending on implementation constraints. The first level corresponds to business level attributes that depend on NS as activities to accomplish business tasks. This is followed by the application system level that applies NS as business activities so that the business activity’s relationship to the application system is clarified. This level transforms the physical service to an abstract service and vice versa. This is carried out in order to match physical business processes to abstract application activities. The last level describes the information technology (IT) infrastructure based on NS constraints in order to describe how well the IT infrastructure would perform when required to deal with the SOA EI system. The mathematical model used to describe reliability for business processes used by Zeng, Huang and Fan (2009) has been expressed as [16]: Where: is the business reliability such that is the value of an indicator promised by a service provider is the actual real value of the particular indicator determined from agreed assessment procedures is the number of consumers involved in the overall assessment On the other hand, the mathematical model used to describe performance for business processes used by Zeng, Huang and Fan (2009) relies on the system response time (). The mathematical relationship can be expressed as [16]: Where: represents the time period an application requires to process an input into an output IV. Simulation for Reliability and Performance in SOAs In order to simulate the business processes using the mathematical relationships described above, a multi-tier inventory system was chosen that would operate through a SOA framework. The inventory system is typical to a number of businesses such that k consecutive layers can describe the inventory system. Each layer in the SOA implementation contains a number of different inventory nodes. Any random node on any layer can be described by (j,k) where j represents the jth node on the kth layer. In the current multi-tier inventory framework, only neighboring layers are allowed to have supply order connections. Moreover, supply order connections can only be created from the k+1th layer to the kth layer. The overall connectivity and tier system is shown in Figure 1 below. The simulation was carried out using an AMD machine with Turion 64 x 2 processing rated at 2 Gigahertz with 3 GB of RAM. IBM INNOV8 was used in order to simulate the mathematical models for reliability and performance. Figure 1 - Multi-tier inventory system depending on a SOA framework V. Results and Findings The results shown below clearly indicate that reliability is low at the start (steps one and two) but soon it gains a constant level at 0.6 after which it tends to drop at the last business processes (between seven and ten) for inventory control. The last step sees some resurgence in reliability but this level is still below the highest achieved level. It must be noticed that the maximum reliability achieved in this simulation is only 0.576 which indicates a low reliability for the overall business situation. This in itself would amount to a loss of over 40% in inventory processing which is unacceptable for any physical business inventory control scheme. This indicates that the reliability offered by the current SOA framework could improve. The low reliability offered by the current model could also be attributed to the simple mathematical model in use that tends to overestimate other factors that are used in physical business systems for redundant error control. Moreover, a lack of operator familiarity could also be cited as a probable cause for low reliability though it needs further investigation before a final verdict could be delivered. This paper proposes that the current mathematical model for reliability could be improved by implementing a model of the form: Where: is the correction factor applied from node to node Alternatively a singular correction factor could also be used for simpler processes or for processes that have a low amount of nodes. Figure 2 - Reliability against Busines Activities for Inventory Control On the other hand, the response time results display low response times shooting into higher response times and then overshooting before settling again for a low response time. This trend can be seen as being analogous to the reliability curve above without the constant region between steps three and six. Although this fails to make sense at first sight but this accurately simulates physical business processes. During inventory control, steps such as purchasing and supply from outside vendors is beyond the control of the business that can be attributed to the increased response times. In contrast, the decrease in response time can be attributed to tighter internal business control. This stands to indicate that the current mathematical model for response time (or alternatively business performance measurement) provides a realistic picture and requires little tuning for actual implementation. Figure 3 - Response Time against Busines Activities for Inventory Control VI. Conclusion The findings of this study indicate that the models presented by Zeng, Huang and Fan (2009) present an inaccurate reliability mathematical model. In contrast, their performance model is more accurate and requires little attention for actual implementation. It is proposed that the reliability model be imbued with a correction factor that could be applied node to node for more complex business processes and as a singular average weighted factor for simpler business processes. The use of a singular weighted correction factor would reduce the overall computational time in contrast to multi-node correction factors. Acknowledgements { acknowledgements } References x [1] Daniel A. Menasce, Honglei Ruan, and Hassan Gomaa, "QOS management in service-oriented architectures," Performance Evaluation, vol. 64, pp. 646-663, 2007. [2] Michael Bell, Service-Oriented Modeling: Service Analysis, Design, and Architecture. New York: Wiley & Sons, 2008. [3] Rashid Al-Ali et al., "A model for Quality of Service provision in Service Oriented Architechtures," International Journal of Grid and Utility Computing, pp. 1-10, 2005. [4] R. J. Al-Ali, A. Hafid, O. F. Rana, and D. W. Walker, "QoS adaptation in service oriented grids," in Process 1st International Workshop Middleware for Grid Computing, Rio de Janeiro, Brazil, 2003. [5] Chunlin Li and LaYuan Li, "A multi-agent-based model for service-oriented interaction in a mobile grid computing environment," Pervasive and Mobile Computing 7(2), pp. 270-284, 2011. [6] C. Cavanaugh, L. R. Welch, B. Shirazi, E. Huh, and S. Anwar, "Quality of service negotiation for distributed, dynaic real-time systems," in Workshop on Bio-Inspired Solutions to Prallel Processing Problems, Fort Lauderdale, Florida, 2002, pp. 757-765. [7] Glen Dobson. (2004) Quality of Service in Service-Oriented Architectures. [Online]. http://digs.sourceforge.net/papers/qos.html [8] M. Bennani and D. A. Menasce, "Resource allocation for autonomic data centers using analytic performance models," in Process Second International Workshop on Bio-Inspired Solutions to Parallel Processing Problems, Seattle, Washington, 2005. [9] D. A. Menasce, "Automatic QoS control," IEEE Computing, vol. 7, no. 1, pp. 92-95, 2003. [10] Lu Liu et al., "Modelling and simulation of Network Enabled Capability on service-oriented architecture," Simulation Modelling Practice and Theory 17(8), pp. 1430-1442, 2009. [11] Jong Myoung Ko, Chang Ouk Kim, and Ick-Hyun Kwon, "Quality of service oriented web service composition algorithm and planning architecture," The Journal of Systems and Software, vol. 81, pp. 2079-2090, 2008. [12] D. Berardi, D. Calvanese, G. De Giacomo, M. Lenzerini, and M. Mecella, "Automatic services composition based on behavioral descriptions," International Journal of Cooperative Information Systems, vol. 14, no. 4, pp. 333-376, 2005. [13] J. Cardoso, "Quality of service and semantic composition of workflows," University of Georgia, PhD Thesis 2002. [14] Michael Bell, SOA Modeling Patterns for Service-Oriented Discovery and Analysis. New York: Wiley & Sons, 2010. [15] H. Zhuge and J. Liu, "Flexible retrieval of web services," Journal of Systems and Software, vol. 70, no. 1-2, pp. 107-116, 2004. [16] Sen Zeng, Shuangxi Huang, and Yushun Fan, "Service-Oriented Enterprise Network Performance Analysis," Tsinghua Science and Technology 14(4), pp. 492-503, 2009. x Read More