Skip to content
FeatureSlurmVelda
Containerized Execution
Each task runs in a container that is isolated with other workloads on the same node, and developers cannot access tasks of other users
Environment customization
Every workload can run in fully customizable environment, including system packages(apt, pip, etc.)
Email
Get notified when your job is completed
Cluster autoscale
Allocate compute resources from Cloud / Kubernetes based on demand
Interactive Development
Developers have access to dedicated dev-environments, with full capability like IDEs and docker access
Service deployment
Run service directly in the cluster in one command, and access them with multiple options like DNS names, port-forwarding or load-balancer
SSO integration
Use your SSO provider to manage user access / login
RBAC
Role based access control: Limit who can see other's job/data
API access
Run jobs, manage workloads, analyze task results, through powerful and secure rest API
Web access
From browser, directly start development in VS Code Web, or view task logs
Third-party

Velda vs. Slurm

What's right for me?

Both Velda and Slurm serve as options for managing HPC clusters. They provide similar command-line interface to run jobs, assuming all nodes have a consistent environment setup.

Architecture and Deployment Model

Slurm's Traditional Node-Based System: Slurm operates on a legacy HPC model where entire compute nodes are allocated to jobs, regardless of actual resource requirements. This approach, designed for traditional MPI applications, creates significant inefficiencies in modern heterogeneous workloads. Users must understand complex partition systems, quality-of-service (QoS) configurations, and resource limit hierarchies.

Velda's Cloud-Native Architecture: Velda leverages multiple backend options to allocate compute resources: VMs from your cloud provider, Kubernetes to deploy workloads as pods, or through custom scripts like SSH. This enables precise resource allocation and efficient multi-tenancy. The platform abstracts away infrastructure complexity while providing enterprise-grade security, automatic scaling, and comprehensive observability.

Security and Isolation

Slurm provides basic process isolation through Linux user accounts and relies on shared filesystems for data access. This approach creates potential security vulnerabilities and makes it difficult to ensure complete workload separation, especially in multi-tenant research environments or when workloads need different system dependencies.

Velda implements per-user-per-instance containers. Each user's workload runs in a completely isolated environment with guaranteed resource limits and secure communication channels, ensuring consistency across all jobs initiated by that instance. This architecture allows fully customizable packages to be installed, while enabling safe multi-tenancy even for sensitive computational workloads.

Developer Experience and Workflow Integration

Traditional HPC systems require significant domain expertise, with users needing to understand batch schedulers, module systems, and job script creation. Debugging running jobs is challenging, and interactive development workflows are poorly supported.

Velda prioritizes developer experience with intuitive command-line interfaces that feel like local development. Full support for modern IDEs, real-time debugging, and seamless port forwarding enables productive development workflows. Teams can use their existing tools and practices without HPC-specific training.

Enterprise Features and Compliance

Modern organizations require robust access controls, audit logging, and compliance capabilities. Slurm's basic accounting and limited RBAC features may not meet enterprise security requirements.

Velda provides comprehensive enterprise features including SSO integration, detailed audit trails, usage reporting, RBAC policies. Integration with existing security infrastructure ensures seamless adoption in regulated environments.

Begin your Velda journey today

Get Started