HYDRA: Decentralized Location-Aware Orchestration of Containerized Applications

The edge computing paradigm, spurred by the Internet-of-Things, poses new requirements and challenges for distributed application deployment. There is a need for an orchestrator design that leverages characteristics that enable this new paradigm. We present HYDRA, a decentralized and distributed orchestrator for containerized microservice applications. This orchestrator focuses on scalability and resiliency to enable the global manageability of cloud and edge environments. It can manage heterogeneous resources across geographical locations and provide robust application control. Further, HYDRA enables the location-aware deployment of microservice applications via containerization. Thus, an application's services may be deployed to separate locations according to expected needs. In this article, the experiments show the orchestrator scaling to 20000 nodes and simultaneously deploying 30000 applications. Further, empirical results show that location-aware application deployment does not hinder HYDRA's performance, and the random resource search algorithm currently being employed may be used as a baseline to find resources in this decentralized orchestrator. Therefore, we conclude that HYDRA is a viable orchestrator design for the new computing paradigm.

Automated summary

This article introduces HYDRA, a decentralized and distributed orchestrator for containerized microservice applications. HYDRA focuses on scalability and resiliency, enabling global manageability of cloud and edge environments. The experiments prove the viability of this orchestrator design.