Integrating and Accelerating Content Classification and Management at P2P Gateways

• Main Authors: Tsao-Jiang Chang, 張朝江
• Other Authors: Ying-Dar Lin
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/11964640827473109156

碩士 === 國立交通大學 === 資訊科學與工程研究所 === 94 === Peer-to-peer (P2P) software runs over dynamic ports in order to disguise their existence. Conventional port-redirecting proxy architecture cannot manage P2P traffic effectively. A gateway-based P2P administration (P2PADM) architecture had been proposed for managing P2P traffic by inspecting content at the application layer. P2PADM executes connection classification in the kernel space and connection management in the user space. Context switch and data passing from the kernel space to the user space is necessary. We propose a new architecture called kP2PADM for improving the performance of P2PADM by moving the P2PADM package from the user space to the kernel space. The main challenge in this work is how to move the code to the kernel space compatibly with Linux kernel without panicking the kernel. The external benchmarking reveals that the throughput of kP2PADM is 88.85 Mbps higher than that of P2PADM. This work also addresses two weaknesses of P2PADM: (1) reconnection issue and (2) redundant delay due to packet loss. Two solutions are proposed for the two weaknesses: (1) connection cache and (2) fast pass.