GeoEkuiper: A Cloud-Cooperated Geospatial Edge Stream Processing Engine for Resource-Constrained IoT Devices With Higher Throughput

Internet-of-things and big data are important technical means to enhance human resilience to natural hazards in Earth Science. Continuous geospatial data streams generated from in-situ monitoring devices provide the first line of defense. An efficient and high-throughput stream processing engine (SPE) is an indispensable infrastructure. It remains a challenge for cloud-based distributed computing frameworks to analyze and process such high-speed and large-volume data in real time. This challenge has driven growing interest in various edge-based SPEs such as EdgeWise and DART to address this issue. There is an urgent need to develop a general-purpose edge SPE with a unique focus on handling the spatial big data challenge on resource-constrained distributed devices. This work introduces a cloud-cooperated geospatial SPE named GeoEkuiper for these devices. It abstracts all operations of spatial data into functions of expressions of Open Geospatial Consortium (OGC) compliant structured query language (SQL) statements and provides a rule-oriented interface to facilitate chaining complex SQL processing logic. User-defined spatial analysis can be easily expressed as a plugin-able extension. It incorporates a cloud native messaging system to transfer rules to other edge or cloud nodes. By taking advantage of the elasticity and reliability of the messaging system on cloud nodes, GeoEkuiper addresses the throughput challenge of resource-constrained devices with a lightweight scheduling strategy considering runtime metrics. Experimental results show that GeoEkuiper can support efficient spatial queries and processing of large-scale spatial stream data on a Raspberry Pi device with enhanced throughput and provide an effective cross-platform solution to achieve edge intelligence.