Function-as-a-Service (FaaS) is becoming increasingly prevalent, and the number of daily service requests in our system ByteDance has even reached 49.6 billion. These requests exhibit “3H” characteristics:high concurrency,high burstiness, andhigh fluctuation, causing the instance scheduling system to face severe bottlenecks even earlier than the data plane.In this paper, inspired by the idea of on-demand scaling of FaaS applications, we proposeGalaxy, a highly scalable instance scheduling system for FaaS platforms.Galaxydecouples the instance scheduling system into functions and introduces a top controller to dynamically adjust the number of instances for each function based on the request load, effectively eliminating bottlenecks under high concurrent requests. We evaluateGalaxywith real-world workloads and the experimental results show thatGalaxycan complete the scaling of over 89,935 instances within 1 second, excluding cold start overhead.Galaxyincreases the scheduling throughput by 8.6 ~ 12.7× compared to the existing control plane.

Graph neural networks (GNNs), as a rising star in machine learning, are widely used in relational data models and have achieved outstanding performance in graph tasks. GNN continuously takes inspiration from mature models in other domains such as computer vision and natural language processing to motivate the development of graph algorithms. However, due to the various data structures from different domains, the cross-domain migration of models has to go through a long period of disassembly and reconstruction, which may not yield the desired results. To preserve the excellent properties of convolution and optimize the migration process from convolutional neural networks (CNNs) to GNNs, we propose a convolution bridge. The convolution bridge realizes the data alignment from CNN to GNN, so that the CNN-based model can be efficiently migrated to the graph structure model. To demonstrate the effectiveness of our migration strategy, we migrated the inception module and U-Net architecture from CNNs to GNNs, named GraInc and GraU-Net, for the node-level task and the graph-level task, respectively. Experimental results show that GraInc and GraU-Net are highly competitive compared to the current state-of-the-art models, particularly on dense graph datasets.