ERSCHED: An Efficient and Reliable Scheduler for Workflows in Multi-Edge Systems

With the rise of edge computing, many users are turning to hosting and running workflow applications in multi-edge environments. This approach can effectively address issues such as resource limitations, low reliability, and high costs that are often encountered in single-edge environments. However, the heterogeneity of resources in multi-edge environments and the differences among edge service providers (in terms of technical architectures, service types, pricing models, etc.) make scheduling tasks in a multi-edge system quite complex. As a result, effectively utilizing multi-edge resources to efficiently and reliably schedule and execute workflows has become a crucial challenge in multi-edge environments. This paper concentrates on the multi-objective workflow scheduling optimization problem when facing reliability constraints in multi-edge environments. We design an efficient and reliable scheduler, namely ERSCHED, for executing workflows in multi-edge systems, with the goal of optimizing makespan and cost simultaneously under reliability constraints. We implement extensive experiments on both real-world and synthetic workflows to validate the efficacy of our scheduler ERSCHED. Simulation results show that ERSCHED can derive better makespan-cost trade-off fronts with significantly higher HyperVolume than similar peer methods. The average improvement of HyperVolume realized by ERSCHED is 48.7%.