Optimal Cost for Time-Aware Cloud Resource Allocation in Business Process

Abstract

Cloud Computing infrastructures are being increasingly used for running business process activities due to its high performance level and low operating cost. The enterprise QoS requirements are diverse and different resources are offered by Cloud providers in various QoS-based pricing strategies. Furthermore, business process activities are constrained by hard timing constraints and if they are not executed correctly the enterprise will pay penalties costs. Therefore, finding the optimal Cloud resources allocation for a business process becomes a highly challenging problem. While optimizing the Cloud resource allocation cost, it is important to respect activities QoS requirements and temporal constraints and Cloud pricing strategies constraints. The aim of the present paper is to offer a method that assists users finding the optimal pricing strategy for Cloud resource used by business process activities. Basically, we use a binary/(0-1) linear program with an objective function under a set of constraints. In order to show its feasibility, our approach has been implemented and the results of our experiments highlight the effectiveness of our proposed solution.

Toward a correct and optimal time-aware cloud resource allocation to business processes

“…In the first allocation r 1 is allocated as a spot instance and in the second allocation as an on-demand instance. We compute the cost for each allocation based on Equation 1that is (i) the sum of the unit hourly price c ijk of r i from pr ij in strategy st ijk by the activity duration d q (we assume that d q = M axD aq ) and (ii) the sum of activity penalty prices [8]. Although a more expensive strategy (on-demand) is used, the BP cost is lower for the second allocation.…”

“…To this end we used timedautomata networks to check BPs' time-constraint behaviors like reachability and deadlock-free [7]. We also used linear programming to optimize the deployment cost of these BPs [8]. Although, due to the complexity of linear programming, handling BPs with large number of activities (200 as per our work in [8]) turned out cumbersome and inefficient.…”

“…We also used linear programming to optimize the deployment cost of these BPs [8]. Although, due to the complexity of linear programming, handling BPs with large number of activities (200 as per our work in [8]) turned out cumbersome and inefficient. In this paper, first, we extend our verification approach presented in [5,6] to check more advanced properties such as liveness to guarantee better correctness of time-constrained, cloud-aware BPs.…”

“…In this paper, first, we extend our verification approach presented in [5,6] to check more advanced properties such as liveness to guarantee better correctness of time-constrained, cloud-aware BPs. Second, we improve our optimization approach presented in [8] to reduce the deployment cost of cloud resources when running more complex and "large" BPs.…”

ViePEP-C: A Container-Based Elastic Process Platform

“…State-of-the-art eBPMS as suggested by, e.g., [7], [8], [9], apply virtual machines (VMs) as computational entities for the execution of elastic processes. However, VMs as heavyweight entities have disadvantages in terms of deployment and start-up time, because of the requirement of an own operating system.…”