我們提出以iWARP為基礎的TCP/IP卸載引擎(TCP/IP Offload Engine, TOE)的設計與實作的方法。iWARP協定介於應用層和TCP層之間，iWARP協定能夠將封包的資料直接放置到目的端應用程式的緩衝區中，即使封包不按照順序到達。目前我們的TOE能夠支援8個UDP服務，4條TCP主動開啟的連線和8條被動開啟的連線同時運作。使用UMC 0.18u的技術，目前的設計可以達到166 MHz的速度，使用130 K個邏輯閘和134 KB 的SRAM作為內部緩衝區。我們估計在全雙工的運作下TOE可以每秒處理2百萬個封包。

We present the design and implementation of an iWARP-based TCP/IP offload engine (TOE). The iWARP protocol is between the application layer and the TCP layer. The iWARP protocol enables the payload of a packet to be placed directly in the destination application’s buffer, even when the packet arrives out of order. The TOE presented in this thesis supports eight UDP services, four TCP active open, and eight TCP passive open connections simultaneously. With the UMC 0.18u technology, the current design achieves 166 MHz in speed, uses about 130 K gates, and 134 KB of SRAM for internal storage. In the full-duplex, we estimate that TOE can process 2 million packets per second.

[1]Srihari Makineni and Ravi Iyer, “Architectural Characterization of TCP/IP Packet Processing on the Pentium® M microprocessor,” in the Proceedings of the 10th International Symposium on High Performance Computer Architecture, HPCA-10, pages 152-161, February 2004.

[2] Jianping Xu, Nitin Borkar, Vasantha Erraguntla, Yatin Hoskote, Tanay Karnik, Sriram Vangal, and Justin Rattner, “A 10Gbps Ethernet TCP/IP processor,” Microprocessor Research, Intel Labs,
available: http://www.hotchips.org/archives/hc15/3_Tue/8.intel.pdf, August 2003.

[3] Juan M. Sola-Sloan, and Isidoro Couvertier, University of Puerto Rico, “A Parallel TCP/IP Offload Engine Framework for a TOE device,” Society for Computer Simulation, 2003.

[4] RDMA Consortium, available: http://www.rdmaconsortium.org/home.

[5] Remote Direct Data Placement Working Group, available: http://www.ietf.org/html.charters/rddp-charter.html.

[6] Brian Hausauer, Chief Architect, and Netffect, Inc, “iWARP Ethernet: Eliminating Overhead in Data Center Designs,” available:
http://www.analogzone.com/nett0403.pdf.

[7] Cheng-Yeh Yu, “Design and Implementation of a Network Offload Engine,” M.S. thesis, Institute of Computer and Communication Engineering, National Cheng-Kung University, 2005.

[8] Behrouz A. Forouzan, Sophia Chung Fegan, ”TCP/IP Protocol Suite,” ISBN 0-07-119962-4.

[9] Information Sciences Institute, University of Southern California, “Transmission Control Protocol Specification,” RFC 793, available: http://rfc.net/rfc793.html, September 1981.

[10] J. Postel, “User Datagram Protocol specification,” RFC 768, available:http://rfc.net/rfc768.html, August 1980.

[11] W. Feng, P. Balaji, C. Baron, L. N. Bhuyan, and D. K. Panda, “Performance Characterization of a 10-Gigabit Ethernet TOE,” in the Proceedings of the 13th Symposium on High Performance Interconnects, pages 58-63, August 2005.

[12] P. Balaji, W. Feng, Q. Gao, R. Noronha, W. Yu and D. K. Panda, “Head-to-TOE Evaluation of High-Performance Sockets over Protocol Offload Engines,” available: http://nowlab.cse.ohio-state.edu/publications/tech-reports/2005/sdp_compare_tr.pdf.

[13] P. Balaji, H. V. Shah and D. K. Panda, “Sockets vs RDMA Interface over 10-Gigabit Networks: An In-depth analysis of the Memory Traffic Bottleneck,” Workshop on Remote Direct Memory Access (RDMA): Applications, Implementations, and Technologies (RAIT 2004) in conjunction with the IEEE Cluster 2004, San Diego, California, September 2004.

[14] C. Reardon, A. George, and C. Cole, "Comparative Performance Analysis of RDMA-Enhanced Ethernet," Workshop on High-Performance Interconnects for Distributed Computing (at HPDC'05), Research Triangle Park, NC, July 2005.

[15] Dennis Dalessandro, and Pete Wyckoff, “ A performance analysis of the Ammasso RDMA enabled ethernet adapter and its iWARP API”, in the Proceedings of RAIT Workshop, Cluster '05, Burlington, MA, September 2005.