LeanBench: comparing software stacks for batch and query processing of IoT data

摘要

Companies produce data in large volumes and in multiple varieties. This trend is intensifying with the deployment of Internet of Things (IoT) devices. Companies need to process data more efficiently and at the edge of the network if they are to remain capable of making timely business decisions based on data. ApacheHadoopandHiveare two widely used data processing systems. However, they rely on complex software stacks that cannot run in a typical IoT gateway device, i.e., a computer with low hardware specifications. An approach to solve this problem is to replace the software with a leaner system with the same functionality. This approach is the value proposition ofUnicage,a data processing system based on Unix shell scripting, that promises better performance by using the operating system directly for execution and inter-process communication.In this paper, we benchmark data processing systems with workloads that compare Unicage with Hadoop and with Hive. The results show that the complexity of the software stack is indeed a significant bottleneck in data processing.