In-memory computing has become an irreversible trend in big data technology, for which the wide popularity of Spark provides a good evidence. Meanwhile, memory storage and management for large data sets are still posing challenges. Out of numerous solutions, Tachyon, a memory-centric distributed storage, well solves the problems faced by many application scenarios. For example, it avoids severe GC issues due to large in-memory data being stored in JVM heap, enables data sharing across applications/jobs through memory, and reduces overhead caused by JVM heap memory allocation and on-heap data management, which help improve the computing engine’s stability and availability. This article uses application examples to demonstrate how to solve real-world big data issues with Tachyon, and to help understand what Tachyon can offer in various big data scenarios.