Application of Dynamic Instance Queuing to Activity Sequences in Cooperative Business Process Scenarios

Abstract

The optimization of their business processes is a crucial challenge for many enterprises. This applies especially for organizations using complex cooperative information systems to support human work, production lines, or computing services. Optimizations can touch different aspects such as costs, throughput times, and quality. Nowadays, improvements in workflows are mostly achieved by restructuring the process model. However, in many applications there is a huge potential for optimizations during runtime as well. This holds particularly true for collaborative processes with critical activities, i.e. activities that require a high setup or changeover time, typically leading to waiting queues in instance processing. What is usually suggested in this situation is to bundle several instances in order to execute them as a batch. How the batching is achieved, however, has been only decided on static rules so far. In this paper, we feature dynamic instance queuing (DIQ) as an approach towards clustering and batching instances based on the current conditions in the process, e.g. attribute values of the instances. Specifically, we extend our previous work on applying DIQ at single activities towards a queuing approach that spans activity sequences (DIQS). The approach is evaluated based on a real-world case study from the manufacturing domain. We discuss limitations and further applications of the DIQ idea, e.g. with respect to collaborative human tasks.

Automatic Resource Allocation in Business Processes: A Systematic Literature Survey

“…Pflug et Manuscript submitted to ACM al. [43] use an event log to cluster similar tasks so that they can jointly be allocated to a resource, and Zhao et al [68] to create a resource allocation heuristic.…”

“…Twenty studies consider workload [2, 4, 5, 18, 19, 21, 26, 27, 34, 35, 43, 45, 52, 62, 64-66, 68, 70, 72]. For workload, availability of a resource is considered most often ( 10studies) [4,26,27,34,35,43,52,64,68,72]. Fourteen studies consider the role attribute [5,7,12,15,22,24,34,35,50,59,62,63,65,69].…”

“…Two works focused on minimizing the process cost [6,29], whereas the other papers set their goal to: finding the best-fitting resource [69], maximizing the throughput [43], minimizing the cycle time [70] or reducing the entropy of worklist entries [66]. Falling outside of these goals is the work by Huang et al [28], which introduces a more generic approach that can be adapted to any goal depending on the concrete use case.…”

“…Some provide approaches for many-to-1 allocations, e.g. [23,43], where the capacity of a resource may be greater than one, and 1-to-many allocations [14], where a team of resources is needed to solve a task. A few approaches support also many-to-many allocations, where a set of resources can work on a set of tasks [47,56].…”