1. [Home](https://www.g2.com/)
2. ...
3. [Cloud Infrastructure Monitoring Software](https://www.g2.com/categories/cloud-infrastructure-monitoring)
4. [Clockwork Discussions](https://www.g2.com/products/clockwork-clockwork/discuss)

[
 ![Clockwork](https://images.g2crowd.com/uploads/product/image/large_detail/large_detail_119653ddd65ed97ed03a08f5e7cb7f05/clockwork-clockwork.png "Clockwork")
](/products/clockwork-clockwork/reviews)

# Clockwork Discussions

Clockwork.io (Clockwork Systems) provides AI fault tolerance, GPU cluster observability, and performance optimization software for distributed AI infrastructure. Clockwork helps AI builders, GPU cloud providers, neoclouds, and enterprise infrastructure teams keep workloads running, troubleshoot performance issues, and reduce wasted GPU-hours across AI training, inference, and reinforcement learning. Clockwork’s software addresses two challenges in large GPU clusters: failures that interrupt workloads and hidden infrastructure problems that slow them down. Its products include: • TorchPass — AI training fault tolerance using live GPU migration and fast snapshots. TorchPass preserves training progress and supports recovery from GPU and node failures, preemptions, and planned maintenance, reducing reliance on costly checkpoint restarts and recomputation. • LinkPass — Network fault tolerance for distributed training and multi-node inference. Delivered as an NCCL network plugin, LinkPass automatically reroutes traffic around failed or flapping links so workloads continue running without changes to model code. • FleetLens — AI fabric observability that connects workload performance to network health. FleetLens helps teams detect congestion, gray failures, and misconfigurations, determine whether the network is causing a slowdown, and pinpoint the affected link, switch, or network interface. • Fleet Audit — Cluster validation that checks software, nodes, cabling, and networks to identify infrastructure issues before and during AI workloads. Clockwork is delivered entirely as software for cloud and on-premises GPU infrastructure, supporting Ethernet/RoCE and InfiniBand networks. Deployment options include protection at the platform layer and integrations with training frameworks such as TorchTitan, Megatron-LM, and DeepSpeed. By reducing disruption, recovery overhead, and troubleshooting time, Clockwork helps teams improve GPU utilization and goodput—the share of GPU time that produces useful work. Learn more at clockwork.io.

## Be the first to start the conversation

Get answers from real professionals in the G2 community.