| Presentation | 2019-09-06 A Study on Fairness of Reinforcement Learning Based Congestion Control Meguru Yamazaki, Miki Yamamoto, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | With fast deployment of high speed wireless access networks, communication environments for Internet accesses have been changing drastically. According to these wide range of network environments, many kinds of TCP congestion control algorithms have been proposed. Each of these TCP versions focuses on specific environment, e.g. wireless loss, and is designed with hard-wired logic, which means there is no universally applicable TCP algorithm. To resolve this technical problem related to hard-wired logic, several machine learning approaches for TCP congestion control has been proposed, e.g. QTCP. In this paper, we show that QTCP has technical problem of unfair condition due to a selfish behavior of machine learning approach. We propose a new QTCP algorithm which is based on AIMD(Additive Increase and Multiplicative Decrease). Our performance evaluation results show that our proposed improvement for QTCP shows good fairness behavior without degradation of throughput or queue length characteristics. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Congestion Control / Fairness / Reinforcemcent Learning |
| Paper # | NS2019-91 |
| Date of Issue | 2019-08-29 (NS) |
| Conference Information | |
| Committee | NS / IN / CS |
|---|---|
| Conference Date | 2019/9/5(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Research Institute of Electrical Communication, Tohoku Univ. |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Session management (SIP/IMS), Interoperability/Standardization, NGN/NwGN/Future networks, Cloud/Data center networks, SDN (OpenFlow, etc.)/NFV, IPv6, Machine learning, etc. |
| Chair | Yoshikatsu Okazaki(NTT) / Takuji Kishida(NTT-AT) / Hidenori Nakazato(Waseda Univ.) |
| Vice Chair | Akihiro Nakao(Univ. of Tokyo) / Kenji Ishida(Hiroshima City Univ.) / Jun Terada(NTT) |
| Secretary | Akihiro Nakao(Osaka Pref Univ.) / Kenji Ishida(NTT) / Jun Terada(NTT Communications) |
| Assistant | Shinya Kawano(NTT) / / Kazutaka Hara(NTT) / Hiroyuki Saito(OKI) |
| Paper Information | |
| Registration To | Technical Committee on Network Systems / Technical Committee on Information Networks / Technical Committee on Communication Systems |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Study on Fairness of Reinforcement Learning Based Congestion Control |
| Sub Title (in English) | |
| Keyword(1) | Congestion Control |
| Keyword(2) | Fairness |
| Keyword(3) | Reinforcemcent Learning |
| 1st Author's Name | Meguru Yamazaki |
| 1st Author's Affiliation | Kansai University(Kansai Univ.) |
| 2nd Author's Name | Miki Yamamoto |
| 2nd Author's Affiliation | Kansai University(Kansai Univ.) |
| Date | 2019-09-06 |
| Paper # | NS2019-91 |
| Volume (vol) | vol.119 |
| Number (no) | NS-194 |
| Page | pp.pp.13-18(NS), |
| #Pages | 6 |
| Date of Issue | 2019-08-29 (NS) |