内嵌海量异构资源和任务调度算法的数据中心电力-算力耦合仿真数据集

随着数据中心用能规模的快速增长，算力资源管理和计算任务调度成为电力–算力协同优化的一个有前景的技术手段。然而，目前仍缺乏能够高效刻画大规模算力调度系统电算协同运行特征的仿真技术。为支撑相关研究与应用，我们构建了一个集成多样化任务调度算法与异构算力资源、覆盖集群–服务器–芯片多层级的仿真平台，并基于此形成了数据中心电力–算力耦合仿真数据集。在构建过程中，数据生成与采集严格遵循科学的数据质量控制流程，确保数据的一致性与可复现性。该数据集整合了任务数据、电价数据、各层级调度策略、能耗特性与多场景仿真结果，全面反映了大规模混部数据中心集群在多资源层级架构、多任务类型及多约束条件下的运行特征。该数据集面向电–算协同研究，支持资源优化调度、需求响应与数据中心能耗管理等问题的建模与验证，可为电力系统与数据中心的协同优化以及新型电力系统的建设提供关键的数据支撑与实验基础。数据集包含内嵌海量异构资源和任务调度算法的数据中心电力－算力耦合仿真数据集－支撑数据1个文件夹，其中：(1)数据中心集群级别的调度数据，格式为.xlxs，约0.02MB；(2)数据中心集群级别的调度算例分析，格式为.docx，约0.16MB；(3)服务器级别的调度数据，格式为.xlxs，约43.30MB；(4)服务器级别的调度算例分析为.docx，约0.28MB；(5)芯片级别的调度数据，格式为.xlxs，约0.46MB；(6)芯片级别的调度算例分析，格式为.docx，约0.33MB。

With the rapid expansion of energy consumption scales in data centers, computing resource management and computational task scheduling have emerged as promising technical approaches for power-computing collaborative optimization. However, there remains a lack of simulation technologies that can efficiently characterize the collaborative operation characteristics of power and computing in large-scale computing scheduling systems. To support relevant research and applications, we constructed a simulation platform integrating diverse task scheduling algorithms and heterogeneous computing resources, covering multi-level architectures from clusters, servers to chips, and developed a power-computing coupled simulation dataset for data centers based on this platform. During the construction process, data generation and collection strictly followed scientific data quality control procedures to ensure data consistency and reproducibility. This dataset integrates task data, electricity price data, scheduling strategies at all levels, energy consumption characteristics and multi-scenario simulation results, comprehensively reflecting the operational characteristics of large-scale co-located data center clusters under multi-resource hierarchical architectures, multiple task types and multiple constraint conditions. Targeting power-computing collaborative research, this dataset supports the modeling and verification of issues such as optimal resource scheduling, demand response and data center energy consumption management, and provides key data support and experimental foundation for the collaborative optimization of power systems and data centers as well as the construction of new-type power systems. The dataset includes one folder named Data Center Power-Computing Coupled Simulation Dataset - Supporting Data, which integrates massive heterogeneous resources and task scheduling algorithms, and contains the following contents: (1) Cluster-level scheduling data of data centers, in .xlsx format, with a size of approximately 0.02 MB; (2) Cluster-level scheduling case analysis of data centers, in .docx format, with a size of approximately 0.16 MB; (3) Server-level scheduling data, in .xlsx format, with a size of approximately 43.30 MB; (4) Server-level scheduling case analysis, in .docx format, with a size of approximately 0.28 MB; (5) Chip-level scheduling data, in .xlsx format, with a size of approximately 0.46 MB; (6) Chip-level scheduling case analysis, in .docx format, with a size of approximately 0.33 MB.

搜集汇总

背景与挑战

背景概述

该数据集是一个包含数据中心电力-算力耦合仿真的数据集，涵盖了集群、服务器和芯片三个层级的调度数据和算例分析，总数据量为44.63MB，包含6个文件。数据集旨在支持电力-算力协同研究，为资源优化调度和能耗管理等问题提供关键数据支撑。