Network-aware compute and memory allocation in optically composable data centers with deep reinforcement learning and graph neural networks

Abstract

Composable data center architectures promise a means of pooling resources remotely within data centers, allowing for both more flexibility and resource efficiency underlying the increasingly important infrastructure-as-a-service business. This can be accomplished by means of using an optically circuit switched backbone in the data center network (DCN), providing the required bandwidth and latency guarantees to ensure reliable performance when applications are run across non-local resource pools. However, resource allocation in this scenario requires both server-level and network-level resources to be co-allocated to requests. The online nature and underlying combinatorial complexity of this problem, alongside the typical scale of DCN topologies, make exact solutions impossible and heuristic-based solutions sub-optimal or non-intuitive to design. We demonstrate that deep reinforcement learning, where the policy is modeled by a graph neural network, can be used to learn effective network-aware and topologically scalable allocation policies end-to-end. Compared to state-of-the-art heuristics for network-aware resource allocation, the method achieves up to a 20% higher acceptance ratio, can achieve the same acceptance ratio as the best performing heuristic with 3 × less networking resources available, and can maintain all-around performance when directly applied (with no further training) to DCN topologies with 10 2 × more servers than the topologies seen during training.