The invention provides an embedded file system-based reading method for streaming data. The method comprises the following steps: receiving a request for reading the streaming data, creating a reading task for the request when the requested streaming data exist in a disk, distributing storage space for the new reading task and initializing relevant parameters; decomposing the reading task into a plurality of subtasks, wherein each subtask is responsible for reading a piece of physically continuous data, and caching the data; fetching the data from subtask cache, performing encapsulation according to the format of the streaming data, submitting to a calling party of the reading task after a piece of data is encapsulated each time, and releasing the subtask after data submission is competed and triggering the next subtask; when all the subtasks are successfully finished, reporting that the tasks are normally finished to the task calling party, and waiting the task calling party to end the current reading task. According to the embedded file system-based reading method provided by the invention, high-throughput and high-concurrence reading of the streaming data is facilitated, and the problem that the concurrence performance of a server is low when a large number of users visit hot data is effectively solved.