In the big data environment, it is inevitable to deal with PB-level and even EB-level data, with magnitudes of 1015 or even 1018. Big data analytics is often prohibitively costly and beyond the reach of most small and medium-sized companies. We strive to rebuild a query processing framework for big data with constrained resources and enable small and medium-sized enterprises to enjoy the real convenience of big data.
Most computing methods do not access all data when small data set within answers. Based on access constraint, we are building a Bounded Evaluation model and theory, by turning big data computing into small data processing. We also design recognition methods of small data aiming at different computing problems. Working on automated data mining and dynamic maintenance algorithm, we study Bounded Evaluation model based on non-row database storage. Based on Bounded Evaluation, the big data real-time analysis platform BEAS, proven by scenario tests, can improve the efficiency of communication data analysis by 25 times to 100,000 times (5 orders of magnitude) and significantly saves computing resources.