Green or Fast? Learning to Balance Cold Starts and Idle Carbon in Serverless Computing
This addresses the challenge of balancing performance and environmental impact for cloud providers and users in serverless computing, representing a novel integration of latency and carbon efficiency.
The paper tackles the trade-off between cold-start latency and carbon emissions in serverless computing by introducing LACE-RL, a framework that uses deep reinforcement learning to dynamically manage pod retention, resulting in a 51.69% reduction in cold starts and a 77.08% decrease in idle carbon emissions compared to a static policy.
Serverless computing simplifies cloud deployment but introduces new challenges in managing service latency and carbon emissions. Reducing cold-start latency requires retaining warm function instances, while minimizing carbon emissions favors reclaiming idle resources. This balance is further complicated by time-varying grid carbon intensity and varying workload patterns, under which static keep-alive policies are inefficient. We present LACE-RL, a latency-aware and carbon-efficient management framework that formulates serverless pod retention as a sequential decision problem. LACE-RL uses deep reinforcement learning to dynamically tune keep-alive durations, jointly modeling cold-start probability, function-specific latency costs, and real-time carbon intensity. Using the Huawei Public Cloud Trace, we show that LACE-RL reduces cold starts by 51.69% and idle keep-alive carbon emissions by 77.08% compared to Huawei's static policy, while achieving better latency-carbon trade-offs than state-of-the-art heuristic and single-objective baselines, approaching Oracle performance.