| 基于本地收益最大化的命名网内计算部署机制 |
| |
| 引用本文： | 田建业,朱轶. 基于本地收益最大化的命名网内计算部署机制[J]. 重庆邮电大学学报(自然科学版), 2024, 36(4): 756-764 |
| |
| 作者姓名： | 田建业 朱轶 |
| |
| 作者单位： | 江苏大学 计算机科学与通信工程学院, 江苏 镇江 212013 |
| |
| 基金项目： | 国家自然科学基金项目(62276116);江苏省教育厅未来网络科研基金项目(FNSRFP-2021-YB-49) |
| |
| 摘 要： | 命名网内计算(named in-network computing, NINC)是一种基于命名数据网络架构、在网络设备处提供泛在计算服务的新兴计算方案。如何有效在单个路由器上部署多个NINC服务,仍有待探索。针对这一问题,提出了本地部署收益与本地服务收益的概念,前者表征NINC服务本地部署后降低转发流量所获得的CPU资源节约量,后者表征所部署的NINC服务每消耗单位CPU资源所获得的流量处理能力。在此基础上,设计了一种基于本地收益最大化的NINC部署机制。该机制中,路由器周期性预评估拟部署NINC服务的本地部署收益与本地服务收益,进而将部署问题转化为一个背包问题,求解出优化的NINC服务部署方案。仿真结果表明,相较于基于流行度的机制,该机制能够有效提高网内计算流量处理能力,在网内计算服务所需数据量均匀分布的一般性场景下,单个路由器可获得约20%~27%的处理能力提升。
|
| 关 键 词： | 命名数据网络 网内计算 软件路由器 优化部署 收益最大化 |
| 收稿时间： | 2023-04-17 |
| 修稿时间： | 2024-02-29 |
Named in-network computing deployment scheme based on maximum local benefit |
| |
| TIAN Jianye,ZHU Yi. Named in-network computing deployment scheme based on maximum local benefit[J]. Journal of Chongqing University of Posts and Telecommunications, 2024, 36(4): 756-764 |
| |
| Authors: | TIAN Jianye ZHU Yi |
| |
| Affiliation: | School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, P. R. China |
| |
| Abstract: | Named in-network computing (NINC) is an emerging computing technology based on named data networking (NDN), which can provide ubiquitous computing services by network devices. However, how to effectively deploy multiple NINC services to achieve better network performance is still a challenge. Aiming at this problem, we propose the concepts of local deployment benefit and local service benefit in this paper. The former represents the saved CPU resource achieved by reducing part of forwarding traffic after NINC is deployed locally. The latter represents the traffic processing capability per unit of CPU resource consumed by deployed NINC. Then, to maximize local resource usage, we further design a NINC service deployment scheme based on maximum local benefit. In this scheme, the NDN software router will first periodically evaluate the local deployment benefit and local service benefit of each selected popular NINC service, and then model the deployment problem as a knapsack problem. Through solving the knapsack problem, an optimal NINC services deployment list is determined. The simulation results show that, compared with the popularity-based scheme, our scheme can effectively improve the capability of processing computing traffic in the network. Under the general scenario where the amount of data required by the in-network computing service follows a uniform distribution, a single router can achieve an increase in processing capacity of about 20% to 27%. |
| |
| Keywords: | named data networking in-network computing software router optimal deployment benefit maximization |
|
| 点击此处可从《重庆邮电大学学报(自然科学版)》浏览原始摘要信息 |
|
点击此处可从《重庆邮电大学学报(自然科学版)》下载全文 |
|
