Training large Deep Neural Network (DNN) models requires thousands of GPUs over the course of several days or weeks. At this scale, failures are frequent and can have a big impact on training throughput. Utilizing spare GPU servers to mitigate performance loss becomes increasingly costly as model sizes grow. ReCycle is a system designed for efficient DNN training in the presence of failures, without relying on spare servers. It exploits the inherent functional redundancy in distributed training systems - where servers across data-parallel groups store the same model parameters - and pipeline schedule bubbles within each data-parallel group. When servers fails, ReCycle dynamically re-routes microbatches to data-parallel peers, allowing for uninterrupted training despite multiple failures. However, this re-routing can create imbalances across pipeline stages, leading to reduced training throughput. To address this, ReCycle introduces two key optimizations that ensure re-routed microbatches are processed within the original pipeline schedule's bubbles. First, it decouples the backward pass into two phases: one for computing gradients for the input and another for calculating gradients for the parameters. Second, it avoids synchronization across pipeline stages by staggering the optimizer step. Together, these optimizations enable adaptive pipeline schedules that minimize or even eliminate training throughput degradation during failures. We describe a prototype for ReCycle and show that it achieves high training throughput under multiple failures, outperforming recent proposals for fault-tolerant training such as Oobleck and Bamboo by up to 1.46× and 1.64×, respectively.

Cluster managers play a crucial role in data centers by distributing workloads among infrastructure resources. Declarative Cluster Management (DCM) is a new cluster management architecture that enables users to express placement policies declaratively using SQL-like queries. This paper presents our experiences in scaling this architecture from moderate-sized enterprise clusters (102- 103nodes) to hyperscale clusters (104nodes) via query optimization techniques. First, we formally specify the syntax and semantics of DCM's declarative language, C-SQL, a SQL variant used to express constraint optimization problems. We showcase how constraints on the desired state of the cluster system can be succinctly represented as C-SQL programs, and how query optimization techniques like incremental view maintenance and predicate pushdown can enhance the execution of C-SQL programs. We evaluate the effectiveness of our optimizations through a case study of building Kubernetes schedulers using C-SQL. Our optimizations demonstrated an almost 3000× speed up in database latency and reduced the size of optimization problems by as much as 1/300 of the original, without affecting the quality of the scheduling solutions.

This paper lays out the rationale for building a completely new operating system (OS) stack. Rather than build on a single node OS together with separate cluster schedulers, distributed filesystems, and network managers, we argue that a distributed transactional DBMS should be the basis for a scalable cluster OS. We show herein that such a database OS (DBOS) can do scheduling, file management, and inter-process communication with competitive performance to existing systems. In addition, significantly better analytics can be provided as well as a dramatic reduction in code complexity through implementing OS services as standard database queries, while implementing low-latency transactions and high availability only once.