상세정보
학술대회
Multi-Agent Learning-based Package Caching in Serverless Edge Computing
- 발행일
- 202210
- 출처
- International Conference on Information and Communication Technology Convergence (ICTC) 2022, pp.400-402
- 협약과제
- 22HH8400, 주문형 데이터 기반 네트워크 지능화 프레임워크 기술 개발, 윤승현
- 초록
- In this paper, we investigate a learning-based caching policy in serverless edge computing (SEC), especially where critical packages for container are cached and shared. To this end, we formulate a cooperative multi-agent decision problem and solve it using deep multi-agent reinforcement learning (DMARL) approach where homogenous agents share a common reward, cache hit rate and QoS violations. The simulation results indicate that the DMARL approach performs better than baselines in the QoS violation rate.
- KSP 제안 키워드
- Cache hit rate, Caching Policy, Cooperative multi-agent, Decision problem, Learning-based, Multi-agent Learning, Reinforcement Learning(RL), Violation rate, edge computing, multi-agent reinforcement learning, simulation results