Approximating Aggregated SQL Queries With LSTM Networks.

摘要: Despite continuous investments in data technologies, the latency of querying still poses a significant challenge. Modern analytic solutions require near real-time responsiveness both to make them interactive and support automated processing. Current technologies (Hadoop, Spark, Dataflow) scan dataset execute queries. They focus on providing scalable storage maximize task execution speed. We argue that these fail offer an adequate level interactivity since they depend continual access data. In this paper we present method for query approximation, also known as approximate processing (AQP), reduce need during inference (query calculation), thus enabling rapid tool. use LSTM network learn relationship between queries their results, provide layer predicting results. Our (referred ``Hunch``) produces lightweight which provides high throughput. evaluated our using 12 datasets. The results show predicted queries' with normalized root mean squared error (NRMSE) ranging from approximately 1\% 4\%. Moreover, was able predict up 120,000 second (streamed together), single no more than 2ms.