HPC compute without complexity.
AIDEL InfraHub HPC 2.0 connects research teams to governed CPU and GPU workloads through a single self-hosted portal and an HTCondor scheduling backend.
See the whole lab.
A private HPC fabric with three clear layers: a control plane that decides who can run what, a mature scheduler that shares compute fairly, and a pool of Linux CPU and GPU workers that do the work.
Access. Allocate. Execute.
Access
Accounts and projects
Allocate
CPU / GPU requests
Execute
Isolated job profiles
Your shared compute. One operating view.
Administrators control project access and compute budgets, while researchers get a clear interface for submitting jobs and collecting results.
A unified workspace
One portal for users, queues, projects, images, monitoring and operational tools.
Meaningful resource policy
Set per-job CPU, memory, whole-GPU and walltime ceilings, and reserve monthly GPU-hour budgets.
Visibility across the lifecycle
Follow submitted, queued, running, held and finished jobs, and review authorized outputs.
Keep the scheduler mature
Use HTCondor for real job matching and scheduling instead of a home-built dispatcher.
Designed for research operations.
Practical controls for teams sharing expensive infrastructure, from admission to day-to-day visibility.
Project governance
Create projects, approve membership and apply role-based access to resource and result views.
CPU + GPU job requests
Request whole CPU cores, RAM, whole GPUs and walltime; validated specs are passed to HTCondor.
Share & budget policies
Generate HTCondor accounting-group shares and apply active-job caps and GPU-hour admission budgets.
Approved workload images
Use admin-approved, SHA256-pinned container profiles and a fixed Docker-universe submit template.
Queue, results & history
Track job states, cancel your own jobs and inspect bounded output with owner-aware authorization.
Infrastructure tooling
Node monitoring, service checks, alerts, Ollama links and administrator backup utilities.
From script to results in five steps.
Author
Write the job and choose an approved image.
Validate
Specs and budgets are checked.
Queue
HTCondor matches it fairly.
Execute
Runs on a Linux CPU / GPU worker.
Review
See results and history.
Your cluster. Your policies.
InfraHub is a self-hosted application gateway. HTCondor and the Linux execution environment are separate infrastructure components, installed and hardened at your site, and AIDEL can do that setup for you.
Windows agents: monitoring only · GPU jobs: dedicated Linux HTCondor execution points
Included in InfraHub
- Project membership and job admission controls
- CPU / GPU requests, quota and budget logic
- Pinned image approval and safe submit templates
- Job lifecycle tracking and cancellation
- Node monitoring, alerts and backup utilities
Set up at your site
- A live HTCondor pool on dedicated Linux execution points
- Worker isolation: cgroups, containers and GPU isolation
- Single sign-on / MFA, security review and logging to your policy
- High availability, recovery objectives and backup checks
- Acceptance testing on your own hardware before go-live
Recommended start: pilot on a private network with a dedicated controller, authenticated HTCondor daemons and Linux Docker workers. Verify a CPU job, then a real NVIDIA GPU job, before onboarding student workloads.
Plan your InfraHub pilot
We'll help you set up the controller, HTCondor and Linux GPU workers, then run your first real jobs together.