ECHO: An Adaptive Orchestration Platform for Hybrid Dataflows across Cloud and Edge

摘要

The Internet of Things (IoT) is offering unprecedented observational datathat are used for managing Smart City utilities. Edge and Fog gateway devicesare an integral part of IoT deployments to acquire real-time data and enactcontrols. Recently, Edge-computing is emerging as first-class paradigm tocomplement Cloud-centric analytics. But a key limitation is the lack of aplatform-as-a-service for applications spanning Edge and Cloud. Here, wepropose ECHO, an orchestration platform for dataflows across distributedresources. ECHO's hybrid dataflow composition can operate on diverse datamodels -- streams, micro-batches and files, and interface with native runtimeengines like TensorFlow and Storm to execute them. It manages the application'slifecycle, including container-based deployment and a registry for statemanagement. ECHO can schedule the dataflow on different Edge, Fog and Cloudresources, and also perform dynamic task migration between resources. Wevalidate the ECHO platform for executing video analytics and sensor streams forSmart Traffic and Smart Utility applications on Raspberry Pi, NVidia TX1, ARM64and Azure Cloud VM resources, and present our results.