Takahiro Suzuki;Sang-Yuep Kim;Jun-ichi Kani;Tomoaki Yoshida
{"title":"Fully softwarized 10G-EPON by low-latency function chaining in container environments","authors":"Takahiro Suzuki;Sang-Yuep Kim;Jun-ichi Kani;Tomoaki Yoshida","doi":"10.1364/JOCN.532796","DOIUrl":null,"url":null,"abstract":"In order to realize flexible optical access systems supporting various services in access networks, including passive optical networks (PONs), through agile development, network function virtualization (NFV) is being expanded from upper-layer management and control functions to physical-layer (PHY) functions. We have already demonstrated the real-time software implementation of 10G-EPON PHY on a general-purpose server with graphic processing units (GPUs). However, to realize desirable microservice architectures where functions are independently developed in each environment, function chaining with conventional inter-container communication schemes, such as transmission control protocol (TCP), incurs significant latency that hinders real-time physical-layer processing. In addition, no study has introduced the chaining of software-based PHY functions and upper-layer functions. The 64b/66b control functions are also problematic due to the serial processing, which adds control symbols to each frame that is input. This paper proposes low-latency function chaining in container environments along with parallelized 64b/66b implementation by separating bit shift operations of 2-bit header addition for fully softwarized 10G-EPON. In our inter-container transfer proposal, a shared data region is made available to the appropriate application when a processing completion notification flag is set in inter-container shared memory. A fully containerized optical line terminal (OLT) and optical network units (ONUs) are implemented and connected; throughput and round-trip latency are evaluated using standard Ethernet frames. Evaluation results show that our function chaining method offers low latency independent of transfer size and that our parallelized 64b/66b implementation method is 19 times faster than the conventional approach. Our containerized PON platform achieves round-trip latency of approximately 1 ms and theoretical 10G-EPON throughput performance of 8.7 Gb/s.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"17 4","pages":"275-284"},"PeriodicalIF":4.0000,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Optical Communications and Networking","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/10926728/","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}
引用次数: 0
Abstract
In order to realize flexible optical access systems supporting various services in access networks, including passive optical networks (PONs), through agile development, network function virtualization (NFV) is being expanded from upper-layer management and control functions to physical-layer (PHY) functions. We have already demonstrated the real-time software implementation of 10G-EPON PHY on a general-purpose server with graphic processing units (GPUs). However, to realize desirable microservice architectures where functions are independently developed in each environment, function chaining with conventional inter-container communication schemes, such as transmission control protocol (TCP), incurs significant latency that hinders real-time physical-layer processing. In addition, no study has introduced the chaining of software-based PHY functions and upper-layer functions. The 64b/66b control functions are also problematic due to the serial processing, which adds control symbols to each frame that is input. This paper proposes low-latency function chaining in container environments along with parallelized 64b/66b implementation by separating bit shift operations of 2-bit header addition for fully softwarized 10G-EPON. In our inter-container transfer proposal, a shared data region is made available to the appropriate application when a processing completion notification flag is set in inter-container shared memory. A fully containerized optical line terminal (OLT) and optical network units (ONUs) are implemented and connected; throughput and round-trip latency are evaluated using standard Ethernet frames. Evaluation results show that our function chaining method offers low latency independent of transfer size and that our parallelized 64b/66b implementation method is 19 times faster than the conventional approach. Our containerized PON platform achieves round-trip latency of approximately 1 ms and theoretical 10G-EPON throughput performance of 8.7 Gb/s.
期刊介绍:
The scope of the Journal includes advances in the state-of-the-art of optical networking science, technology, and engineering. Both theoretical contributions (including new techniques, concepts, analyses, and economic studies) and practical contributions (including optical networking experiments, prototypes, and new applications) are encouraged. Subareas of interest include the architecture and design of optical networks, optical network survivability and security, software-defined optical networking, elastic optical networks, data and control plane advances, network management related innovation, and optical access networks. Enabling technologies and their applications are suitable topics only if the results are shown to directly impact optical networking beyond simple point-to-point networks.