dstack SSH fleets
AI container orchestration tool
FreeOpen SourceAI1,465 impressions#14 of its week
dstack provides an open-source orchestration layer to manage AI workloads across GPU clouds, Kubernetes, VMs, and bare-metal via SSH fleets.
- for
- AI engineers, data-center operators, and teams using heterogeneous on-prem or cloud GPU compute.
- pricing
- open source
- license
- MPL-2.0
Key features
- SSH Fleets — Connect and provision VMs or bare-metal servers over SSH for AI training and inference.
- Kubernetes Backend — Integrate existing Kubernetes clusters to schedule containerized AI jobs.
- Slurm Backend — Experimental support for Slurm clusters, submitting runs as Slurm jobs.
- Tasks Scheduling — Schedule training and other AI jobs with first-class compute primitives.
- Services — Cache-aware, PD-disaggregated inference serving with auto-scaling.
- Presets — Agent-based optimization toolkit for workload configuration.
- Projects — Tenant isolation and usage metering for multi-tenant AI labs.
- Gateways — HTTPS endpoints with domain, rate-limit, and auto-scaling support.
Use cases
- Spin up multi-node training across on-prem servers and cloud GPUs without manual provisioning.
- Run inference services on heterogeneous clusters with automatic scaling and caching.
- Unify workload management for AI labs that use both Kubernetes and bare-metal resources.
- Replace Slurm or generic Kubernetes orchestration with a lightweight AI-focused interface.
dstack SSH fleets vs alternatives
 dstack SSH fleets |  Daytona |  BuildShip | .png?width=64) Langfuse |  Keel | |
|---|---|---|---|---|---|
| Best for | Unified AI workload orchestration | Dev environment provisioning | Visual backend creation | LLM analytics | App backend tooling |
| Pricing | Open source | Free | Free | Subscription | Subscription |
| DevHunt upvotes | 0 | 339 | 118 | 89 | 46 |
| Launched | Jun 2025 | May 2024 | Jan 2023 | Jan 2023 | Mar 2024 |
- dstack SSH fleets vs Daytona: Daytona focuses on setting up development environments, not AI workload orchestration.
- dstack SSH fleets vs BuildShip: BuildShip is a visual backend builder, whereas dstack orchestrates AI compute across heterogeneous resources.
- dstack SSH fleets vs Langfuse: Langfuse provides LLM observability, while dstack manages compute and job scheduling.
- dstack SSH fleets vs Keel: Keel offers backend services for apps, not a unified AI compute orchestration layer.
dstack SSH fleets FAQ
Can I use dstack with Kubernetes?+
Yes. Connect existing Kubernetes clusters through the Kubernetes backend and dstack will schedule AI workloads alongside other backends.
Can I use dstack with Slurm?+
Yes. The experimental Slurm backend connects to the cluster’s login node over SSH and submits runs as Slurm jobs, requiring Pyxis and enroot.
What is the difference between dstack and Slurm?+
Slurm is a traditional HPC workload manager; dstack adds a unified, container-focused orchestration layer with AI-specific primitives for compute, training, inference, and observability.
What is the difference between dstack and Kubernetes?+
Kubernetes is a general-purpose container orchestrator; dstack provides a lightweight AI-specific interface with first-class compute, training, and inference features.
How do I get support?+
Support is available via Discord or direct contact as listed on the dstack website.
Summarized by DevHunt from dstack.ai · Sep 28, 2026. Details may change; check the official site.
.png?width=64)
