Task allocation method, device and storage medium of operation server

Abstract

The application relates to a task allocation method and device of an operation server and a storage medium, and particularly relates to the field of computer resource scheduling and management, and through construction of a deep joint portrait of task general features and hardware real-time states, precise matching of heterogeneous computing resources and diversified workloads is realized, multi-target benefit decision-making is upgraded from traditional resource allocation, a graph structure is used to fuse network topology and task adaptation relationship, and through context-aware state embedding, a global view is provided for collaborative decision-making, a multi-agent reinforcement learning and shadow deployment closed-loop mechanism are introduced, the system can continuously optimize a prediction model and a scheduling strategy based on real feedback, dynamic trade-off and long-term optimization of energy efficiency, cost and other targets are realized under the premise of meeting service quality, and an intelligent scheduling body with self-learning and self-adaptive capabilities is formed.

Images