Fig. 6

A broadcast cycle of 23 frames based on Hadoop. We assume that the sequence of the frame broadcast starts from the Data Node 1, in which all frames are broadcast firstly, and then frames in the Data Node 2 are sent in order, and so on. All frames (each frame includes index and data item) are divided equally over the Hadoop and become a sequence of frames, which are then stored in the corresponding Data Node. The number of frame sequences stored on each Data Node is ⌈D/n⌉, where D is the number of frames of the broadcast data and n is the number of Data Node in the cloud platform. The primary index structure consists of two parts: the first part is a quadruple, which denotes the index of frame information in each Data Node (Data Node, Number of Frames, Start Frame, End Frame) and the second part is also a quadruple, which denotes the index of frame number belonging to each Partition ID (FN(ID0), FN(ID1), FN(ID2), FN(ID3)). For the first part, parameter Data Node represents server number, in which can store the frames, parameter Number of Frames represents the total number of frames stored in the server, Start Frame represents triple value of the first frame stored in the Data Node, and End Frame represents triple value of the last frame stored in the Data Node. In the second part, parameter FN(ID0) indicates the number of cells in Partition ID0 (namely, the frame number of partition ID0), and so on