KAI Scheduler vs TORQUE Resource Manager
KAI Scheduler
kai-scheduler.devTORQUE Resource Manager
github.comSide by side
| Figure | KAI Scheduler | TORQUE Resource Manager |
|---|---|---|
| Price from | — | Free |
| Free plan | — | Yes |
| Free trial | — | — |
| Open source | — | — |
| macOS | — | Yes |
| Linux | — | Yes |
| API | — | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 0 | 1 |
| Spec score · GPU Cluster Management Software | 7.4 (#13) | 7.5 (#12) |
| Integrations | The project states it is natively integrated with KubeRay for Ray workloads on Kubernetes and describes integrations with Grove and Dynamo for complex multi-component workloads. | The maker's documentation describes integration between TORQUE and the Moab Workload Manager. |
| License | The public GitHub repository identifies the project as open source and lists an Apache-2.0 license. | Use, modification, and distribution are subject to the terms in the repository's PBS_License.txt file. |
| Purpose | KAI Scheduler is a Kubernetes-native scheduler for AI workloads that optimizes GPU allocation across data processing, training, and inference while keeping resource shares fair across teams. | TORQUE is an open-source resource manager and queue manager based on the original PBS resource manager. |
| Support | The project directs users to GitHub issues for bugs, feature requests, or technical help, and also lists CNCF Slack, community calls, and a mailing list. | The project README directs questions and contributions to the [email protected] mailing list. |
Plans
In their own words
KAI Scheduler
KAI Scheduler is a self-hosted, open-source Kubernetes scheduler for allocating GPUs to AI workloads such as data processing, training, and inference. It supports interactive notebooks through multi-node distributed training in one cluster and is designed for large GPU clusters and high-throughput workloads. Hierarchical queue trees provide quotas, limits, priorities, and borrowing between teams. Gang scheduling places distributed workloads together when resources are available, while elastic workloads change replica counts within configured minimum and maximum values. GPU sharing includes time slicing, MPS, and MIG. Other scheduling controls include topology-aware placement, time-based fairshare, and priority-based preemption. KAI documents integrations with Ray on Kubernetes and Grove. It runs on cloud or on-premises infrastructure, including Karpenter environments. Installation uses a production Helm chart, source builds, or ArgoCD, with a running Kubernetes cluster, Helm CLI, and NVIDIA GPU Operator required for GPU requests. The Apache-2.0 project is also a CNCF sandbox project.
TORQUE Resource Manager
TORQUE Resource Manager is an open-source resource and queue manager for compute clusters, based on the original PBS resource manager. The project is free, with Linux, macOS, API and self-hosted options. TORQUE 4.0 release notes state that it can support clusters with more than 15,000 compute nodes; that release also introduced a multithreaded pbs_server intended to improve command response and job throughput. An optional MOM hierarchy can reduce update traffic from compute nodes to pbs_server and lets administrators manage network communication. In version 4.0, component communication moved to TCP/IP, removing UDP and RPP. The trqauthd daemon authorizes client connections to pbs_server and must run as root on client hosts. Linux builds can optionally use control groups and cpusets, and documentation describes integration with Moab Workload Manager. The project is built from source using configure and make; installation instructions also cover RPM and self-extracting package creation. Its license terms are in the repository's PBS_License.txt. A compatibility caveat for version 4.0 is that its components cannot communicate with earlier TORQUE versions, so rolling upgrades will not work.