Research On Method Of High-speed Data Revert Based On HTTP/2

Because of use of binary message framing, multiplexing of requests and server push, HTTP/2 makes network performance promoted.However, it also makes HTTP/1.1 protocol reassembling method not available for HTTP/2. By the following way, this paper implement the research on method of high-speed data revert based on HTTP/2:analyzing the format of the frame; researching on the relationship between frame, stream and connection; toggling the stream state;extracting and storing the message payload.The main research contents include:1 By the following way, researching and implementing the HTTP/2 session management algorithm: analyzing the format of the binary message frame; testing the legitimacy of the frame; toggling the stream state. The HTTP/2 session management module is divided into HTTP/2 frame session management module and HTTP/2 stream session management module. HTTP/2 frame session management module should finish the following work: according to the frame format and the data which is from HTTP/2 starting module, extracting the values of all fields in a frame. HTTP/2 stream session management should finish the following work: according to the length limit of one field in a frame and the dependency among the fields, checking whether the frame is rightful;according to the current stream state and the frame type, toggling the current stream state.2 Through the analysis of the frame payload, researching and implementing the content splicing algorithms. HTTP/2 content splicing module should finish the following work: according to the frame format,extracting the message data from the frame payload in order, updating the dynamic table and storing the promised stream identifier.3 Through the HPACK decompression of the message header field and the decompression and storage of the message payload, HTTP/2 content extraction module should finish the following work: according to the compression format of the message header fields, the current dynamic table, the static table and static Huffman encoding, decompressing the header fields; according to the decompressed message header fields and,storing the message payload.The function is implemented, and the system performance is tested.It can be up to 440Mbps in single thread, but 4300Mbps in twelve threads.