Cloud Execution & Distributed Computing
Run large-scale simulations on the cloud — parametric sweeps, batch processing, and industry-scale models.
1. What is it?
Cloud execution offloads computation from your local machine to a scalable backend:
- Durable job queue: SQLite-backed queue survives restarts — jobs are never lost.
- Content-addressed artifact store: SHA-256 deduplication — same inputs produce the same artifact without recomputation.
- Worker pool: Multiple workers process jobs in parallel with signal handling for graceful shutdown.
- REST API: Submit, poll, cancel, and retrieve results programmatically.
"I need to run 1000 simulations for a parametric study — submit them all and let the cloud handle it."
2. When do I use it?
| Use when | Don't use for |
|---|---|
| Large-scale parametric studies | Single quick simulation (use local solver) |
| Batch processing of many models | Interactive exploration |
| Industry-scale models (400k+ DOF) | Small models that solve in seconds |
| Monte Carlo reliability studies | Single deterministic run |
| Collaborative workflows | Solo work |
3. The flow — every action described
Step 1: Create a job
What you do: Define the simulation to run.
How:
- Click Cloud tab
- New Job → select template or upload model
- Define parameter ranges for sweep (e.g., force = 100..1000 N, step 100)
- Set solver configuration (backend, mesh size, etc.)
Why: The job defines what to run and with what parameters.
Step 2: Submit to queue
What you do: Submit the job to the cloud queue.
How:
- Click Submit
- The job is assigned a job_id
- The queue prioritizes and schedules workers
Why: The job enters the queue and waits for an available worker.
Step 3: Monitor progress
What you do: Track job status and progress.
How:
- Click Jobs tab
- View job list with status (queued, running, completed, failed)
- Click a job to see detailed progress
Why: You can see which jobs are running and which are done.
Step 4: Retrieve results
What you do: Download completed results.
How:
- Click Artifacts tab
- Select completed job
- Download result files (VTK, CSV, JSON)
Why: The artifacts contain the simulation results.
Step 5: Visualize and export
What you do: Visualize the results and export reports.
How:
- Click Results tab
- View contour plots and time histories
- Export PDF report or CSV data
Why: The results are ready for analysis and sharing.
4. Reading your results
Job status
| Status | Meaning |
|---|---|
| queued | Waiting for worker |
| running | Being processed |
| completed | Results available |
| failed | Error occurred |
Artifact metadata
| Field | Meaning |
|---|---|
| job_id | Unique job identifier |
| created_at | Submission time |
| duration | Wall-clock time |
| input_hash | SHA-256 of inputs |
| artifact_size | Size of result data |
5. Boundaries of truth
What's validated
| Case | Reference | Error |
|---|---|---|
| Job queue durability | 22 tests | 0 failures |
| Artifact deduplication | SHA-256 | Exact |
| Worker graceful shutdown | Signal handling | 0 failures |
| REST API contract | 8 tests | 0 failures |
What's NOT covered
- Kubernetes orchestration — manual scaling only
- Redis backend — SQLite only (single-node)
- GPU acceleration — CPU workers only
- Priority preemption — FIFO queue only
- Multi-region deployment — single region only
Common mistakes
| Mistake | Fix |
|---|---|
| Too many parallel jobs | Limit concurrency to worker count |
| Large artifact sizes | Compress results before storing |
| Wrong solver configuration | Verify solver settings before submission |
| No timeout set | Set job timeout to prevent hanging |
| Ignoring job logs | Check logs for convergence issues |
See also
- [Static Structural](static-structural.md) — for single-model analysis
- [Stochastic](stochastic.md) — for Monte Carlo studies
- [Optimization](optimization.md) — for parametric sweeps
Keep exploring
Open the interactive workspace — mesh, solve, validate and export in the browser.
Start a guided solve