从软件分发到部署：优化数据中心效率的全链路探究

doi:10.3778/j.issn.1002-8331.2305-0301

摘要/Abstract

摘要： 针对分布式军事指挥中心的软件分发和部署，进行了软件分发效率提升和软件部署稳定性和资源利用率提升的研究。为了提升软件分发效率，将软件划分为多个切片，同时利用其他指挥中心作为中继节点，通过不同路径并行传输，实现数据的协同传输。针对软件部署，考虑了负载均衡的问题，并利用改进的整数多背包模型构建了部署方案，以提高系统的稳定性和资源利用率，实现系统的稳定性和资源利用率的提升。仿真结果表明，该分发模型相比于传统的直传方式可缩短85%传输时间，软件部署模型相比于软件部署至最空闲服务器的贪婪模型最高可提升30%的资源利用率。分发和部署的实验证明，分发效率可平均缩短50%时间，部署资源利用率可实现接近20%的提升。

关键词: 软件分发, 软件部署, 极小极大整数规划模型, 整数背包模型

Abstract: In the context of distributed military command centers, the research focuses on optimizing software distribution and deployment processes, with an emphasis on improving distribution efficiency and enhancing deployment stability and resource utilization. To enhance software distribution efficiency, a strategy involving the partitioning of software into multiple slices is employed. Concurrently, other command centers are utilized as intermediary relay nodes, enabling parallel transmission through distinct pathways and facilitating collaborative data dissemination. Regarding software deployment, the challenge of load balancing is addressed. Leveraging an enhanced integer multiple knapsack model, a deployment scheme is devised to bolster system stability and optimize resource allocation. This approach seeks to achieve heightened system stability and improves resource utilization. Simulation results indicate that in comparison to traditional direct transmission methods, the proposed distribution model can reduce transmission time by up to 85%. Moreover, the software deployment model, when compared to a greedy approach of deploying software to the least utilized servers, demonstrates a potential resource utilization increase of up to 30%. Empirical validation of the distribution and deployment strategies reveals an average reduction of 50% in distribution time and a potential enhancement of nearly 20% in deployment resource utilization.

Key words: software distribution, software allocation, minimax integer linear programming（ILP） model, integer multi-knapsack model

栾明君, 罗翔, 曹孝元. 从软件分发到部署：优化数据中心效率的全链路探究[J]. 计算机工程与应用, 2023, 59(20): 326-332.

LUAN Mingjun, LUO Xiang, CAO Xiaoyuan. From Software Distribution to Deployment：Exploring Entire Chain of Data Center Efficiency Optimization[J]. Computer Engineering and Applications, 2023, 59(20): 326-332.

参考文献

[1] 戴文博，徐珞，卫津逸.面向军事信息系统的自动化软件部署算法[J].计算机与现代化，2020（1）：90-95.
DAI W B，XU L，WEI J Y.Automation software deployment algorithm for military information system[J].Computer and Modernization，2020（1）：90-95.
[2] CAO X，POPESCU I，CHEN G，et al.Optimal and dynamic virtual datacenter provisioning over metro-embedded datacenters with holistic SDN orchestration[J].Elsevier Optical Switching and Networking，2017，24：1-11.
[3] LIU H，ZHANG Z，LIU Z，et al.Software allocation strategy based on performance and heterogeneity[C]//2020 IEEE 5th Information Technology and Mechatronics Engineering Conference（ITOEC），2020：1092-1096.
[4] ZHAO B，LI Z，ZHANG T.The heterogeneous allocation tool for hardware and software co-design[C]//2020 International Conference on Computer，Information and Telecommunication Systems（CITS），2020：1-5.
[5] WANG H，FENG L，JIN Y，et al.Surrogate-assisted evolutionary multitasking for expensive minimax optimization in multiple scenarios[J].IEEE Computational Intelligence Magazine，2021，16：34-48.
[6] FIORANI M，SAMADI P，SHEN Y，et al.Flexible architecture and control strategy for metro-scale networking of geographically distributed data centers[C]//42nd European Conference on Optical Communication（ECOC），2016.
[7] CHEN G，CAO X，ZHANG D，et al.First demonstration of holistically-organized metro-embedded cloud platform with all-optical interconnections for virtual datacenter provisioning[C]//Opto Electronics and Communications Conference（OECC 2015），2015.
[8] CAO X，LUAN K，LUO X，et al.Accelerating data delivery service with collaborated parallel traffic scheduling for distributed datacenters[C]//2021 IEEE 21st International Conference on Communication Technology（ICCT），2021：306-312.
[9] LUO X，CAO X，LUAN K，et al.Optimal software distribution and allocation strategies for improved datacenter resource utilization[C]//2022 IEEE 22nd International Conference on Communication Technology（ICCT），2022：1868-1872.
[10] AHUJA R.Minimax linear programming problem[J].Operations Research Letters，1985，4：131-134.
[11] HUANG F，GAO S，PEI J，et al.Accelerated zeroth-order and first-order momentum methods from mini to minimax optimization[J].Journal of Machine Learning Research，2022，23：1-70.
[12] THEKUMPARAMPIL K，JAIN P，NETRAPALLI P，et al.Efficient algorithms for smooth minimax optimization[C]//Advances in Neural Information Processing Systems，2019.
[13] AGARWA B，TOGOU A，RUFFINI M，et al.PIRS 3 A：a low complexity multi-knapsack-based approach for user association and resource allocation in HetNets[C]//2021 IEEE International Conference on Communications Workshops（ICC Workshops），2021.
[14] MANCINI S，CIAVOTTA M，MELONI C.The multiple multidimensional knapsack with family-split penalties[J].European Journal of Operational Research，2021，289：987-998.
[15] 戴文博.面向持续软件工程的自动化部署技术研究[D].北京：中国电子科技集团公司电子科学研究院，2020.
DAI W B.Research on automatic deployment technology for continuous software engineering[D].Beijing：Electronic Science Research Institute of China Electronics Technology Group Corporation，2020.