Doctor's Theses (authored and supervised):
"Elastic Business Process Management in the Cloud";
Supervisor, Reviewer: S. Dustdar, R. Steinmetz;
Institut für Informationssysteme, Distributed Systems Group,
oral examination: 2015-10-14.
Business Process Management is a multifaceted approach covering several aspects of organizational, management and technical facets of business processes. Recently, this technology gained great attention in the field of many different industries including the finance industry or the energy domain where computational resources are used to carry out business processes automatically. Business processes are composed from human or software-based services. Especially in larger companies, an extensive number of different processes are available, each made up from various process steps which are realized by
software services, each needing a different amount of resources. As the amount of process instances to be executed varies over time, a Business Process Management System is needed which is able to serve the ever-changing demand of needed resources. For example, the amount of required resources during peak times will be much higher than during off-peak times. Permanent provisioning of resources is obviously not the best choice, as resources which are able to handle peak loads will hardly be used during off-peak times.
Hence, this leads unwanted high cost (i.e., over-provisioning). Contrary, providing less resources may lead to an under-provisioned system, which might not be able to carry out all processes during peak times or will suffer from a low Quality of Service. With
the upcoming of cloud computing, it is possible (i) to lease and release resources in an on-demand, utility-like fashion, and to provide the means of (ii) rapid elasticity through scaling the infrastructure up or down, based on (iii) pay-per-use through metered service. However, so far, only few researchers have provided methods and solutions to facilitate Elastic Processes, i.e., processes which are carried out on cloud-based resources while considering all three dimensions of elasticity: resource, quality and cost.
Hence, this thesis presents novel approaches to apply cloud-based computational resources for Business Process Management. First, we show how leased resources can be used more efficiently if the processī future resource demand is considered during scheduling and resource allocation. Second, by constantly extending this approach,
we present a cost-based optimization model for sequential business processes. As in real-world complex business processes are more realistically, we present afterwards a multi-objective optimization model for complex business processes. Finally, we introduce an extension to this model, allowing to lease resources from a public and private cloud equally, leading to a hybrid cloud environment. We evaluate our approaches extensively and compare the results against state of the art baselines in order to show its potential of reducing process execution time and avoiding unnecessary cost while still ensuring Service Level Agreements.
Created from the Publication Database of the Vienna University of Technology.