Limited by the small on-chip memory, hardware-based transport typically implements go-back-N loss recovery mechanism, which costs very few memory but is well-known to perform inferior even under small packet loss ratio. We present MELO, an efficient selective retransmission mechanism for hardware-based transport, which consumes only a constant small memory regardless of the number of concurrent connections. Specifically, MELO employs an architectural separation between data and meta data storage and uses a shared bits pool allocation mechanism to reduce meta data on-chip memory footprint. By only adding in average 23B extra on-chip states for each connection, MELO achieves up to 14.02x throughput while reduces 99% tail FCT by 3.11x compared with go-back-N under certain loss ratio.
Yuanwei Lu, Guo Chen, Zhenyuan Ruan, Wencong Xiao, Bojie Li, Jiansong Zhang, Yongqiang Xiong, Peng Cheng, Enhong Chen.
Proceedings of the First Asia-Pacific Workshop on Networking (APNet ‘17). [PDF] [Slides]
I only participated in the discussion of the MELO project. Credits should go to Dr. Yuanwei Lu.
- Yuanwei Lu, 4th year Ph.D. student in MSRA and USTC (now in Tencent)
- Prof. Guo Chen, Associate Professor, Hunan University
- Zhenyuan Ruan, 4th year undergraduine in USTC (now a Ph.D. student in UCLA)
- Wencong Xiao, 3rd year Ph.D. student in MSRA and Beihang University
- Bojie Li, 3rd year Ph.D. student in MSRA and USTC
- Dr. Jiansong Zhang, Senior Expert, Alibaba
- Dr. Yongqiang Xiong, Lead Researcher in Microsoft Research Asia
- Dr. Peng Cheng, Researcher in Microsoft Research Asia
- Prof. Enhong Chen, Professor in USTC