Research On Carbon Emission Reduction Regulation Method For Interconnected Data Centers Based On Spatio-temporal Dual-dimensional Load Migration

With the massive global deployment of data centers and the demand surge for cloud computing services,their high energy consumption and carbon pollution is becoming more acute,and it is a significant challenge to mitigate the harmful effects of multi-data center’s carbon footprint.Renewable energy generation,represented by wind and solar power,has opened new opportunities for data centers to reduce emissions.However,renewable energy is highly volatile and uncertain,while data center IT equipment has the stringent power requirements of continuity and stability.This contradiction has always constrained the further use of renewable energy in data centers and prevented the growth of carbon emissions from being radically curbed.This paper proposes an interconnected multi-data center load regulation method to enhance the renewable energy consumption capacity while ensuring the quality of service,effectively reducing carbon emissions.From the perspective of green energy,the widespread global distribution of renewable energy is highly complementary in terms of spatial and temporal distribution;From the perspective of power load,in the spatial dimension,computing loads can be balanced across multi-region through the interconnected network,and in the temporal dimension,delaying or activating computing tasks can fine-grained their short-term power demand.Based on the above source-load features,this paper integrates the data centers’ load composition and the power consumption of interconnected optical networks to build a carbon emission model and characterize their carbon footprint in detail.Carbon optimization models that consider the power consumption of task transfers are constructed by combining the flexibility of computing loads and the spatiotemporal complementarity of renewable energy sources.A spatio-temporal dualdimensional task migration mechanism is proposed to achieve low-carbon operations:A cross-domain spatial migration mechanism for delay-tolerant tasks is established,where the Benders decomposition algorithm is used to solve the task migration scheme and routing selection,effectively solving the proposed mixed-integer programming,and realizing collaborative carbon reduction among multi-data centers;the single data center task temporal migration mechanism is proposed to meet the real-time matching between computing load and renewable energy output.This enables the redistribution of computing loads in an optimized and complementary manner,and tracks renewable energy power fluctuations in real-time,so that data centers’ excess emissions can be shifted and offset using surplus renewable energy from other locations or time points,providing an effective solution to green computing.The experiments demonstrate the effectiveness of the spatio-temporal dualdimensional task migration mechanism.Using the real-world data center workload traces and fine-grained renewable energy output data,the results show that the method in this paper can significantly improve renewable energy consumption capacity and reduce carbon emissions.Under the scenarios with variable parameters,such as renewable energy penetration,resource utilization range and delay-tolerant task ratio,this mechanism can still demonstrate significant emission reduction,verifying the general applicability of the method in complex scenarios.