Inproceedings,
Task Scheduling on Crowdsourcing Platforms for Enabling Completion Time SLAs
31th International Teletraffic Congress (ITC 31), Budapest, Hungary, (2019)
Abstract
Today a diverse crowdsourcing economy has established, and similar to other service industries, SLAs are a crucial part of this ecosystem. However, the uncertainty introduced by human workers makes it harder to estimate and set completion time guarantees than in a system that is solely technology based. In this work, we, therefore, analyze an exemplary task scheduling strategy that enables operators of crowdsourcing-based services to meet completion time SLAs. In particular, we use the real-world crowdsourcing-based text digitalization platform ScaleHub as the foundation for our system model and a dataset provided by ScaleHub to obtain a realistic parametrization of the model. We derive an analytical model that enables us to illustrate the potential of scheduling mechanisms to meet SLA constraints and that helps platform providers to optimize their systems. Further, we use a more detailed simulation model to illustrate the challenges arising from erroneous worker submission in the context of task schedulers.
