Please use this identifier to cite or link to this item:
|Title:||QoS control of E-business systems through performance modelling and estimation|
|Abstract:||E-business systems provide the infrastructure whereby parties interact electronically via business transactions. At peak loads, these systems are susceptible to large volumes of transactions and concurrent users and yet they are expected to maintain adequate performance levels. Over provisioning is an expensive solution. A good alternative is the adaptation of the system, managing and controlling its resources. We address these concerns by presenting a model that allows fast evaluation of performance metrics in terms of measurable or controllable parameters. The model can be used in order to (a) predict the performance of a system under given or assumed loading conditions and (b) to choose the optimal configuration set-up for certain controllable parameters with respect to specified performance measures. Firstly, we analyze the characteristics of E-business systems. This analysis leads to the analytical model, which is sufficiently general to capture the behaviour of a large class of commonly encountered architectures. We propose an approximate solution which is numerically efficient and fast. By mean of simulation, we prove that its accuracy is acceptable over a wide range of system configurations and different load levels. We further evaluate the approximate solution by comparing it to a real-life E-business system. A J2EE application of non-trivial size and complexity is deployed on a 2-tier system composed of the JBoss application server and a database server. We implement an infrastructure fully integrated on the application server, capable of monitoring the E-business system and controlling its configuration parameters. Finally, we use this infrastructure to quantify both the static parameters of the model and the observed performance. The latter are then compared with the metrics predicted by the model, showing that the approximate solution is almost exact in predicting performance and that it assesses the optimal system configuration very accurately.|
|Appears in Collections:||School of Computing Science|
Files in This Item:
|Ferrari G 2007.pdf||Thesis||13.38 MB||Adobe PDF||View/Open|
|dspacelicence.pdf||Licence||43.82 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.