Transwarp Inceptor uses its own high-efficiency columnar storage format and Apache Spark computation engine optimized for in-memory computation, avoiding frequent readings from/writings to harddrives practiced by Map/Reduce framework. Morever, Spark uses a light-weight dispatch framework and a multithreading computation model. As a result, compared to Map/Reduce, Spark has very little dispatch and start-up overhead, higher execution speed and shorter mean time to restoration (MTTR). With respect to real-time online applications, Transwarp Hyperbase has constructed a global index, an auxiliary index and a full-text index. It extends from SQL syntax and can satisfy needs for low latency from online storage and OLAP. With optimizations made on the execution engine and data storage layer, Transwarp Data Hub has overtaken the opensource Apache Hadoop 2.5. TDH has much stronger performance and much more comprehensive SQL support compared to Cloudera Impala. TDH is faster than mainstream MPP databases by 1.5 to 10 times.