# Scaling a real workload

Take a single-machine geospatial pipeline to a cluster, and find out where the time actually goes.

Source: https://university.eugo.io/courses/scaling-a-real-workload

---

- Level: Advanced
- Duration: 38m
- Lessons: 5
- Topics: distributed-compute, data-io, performance

## Lessons

### The starting point

- [The workload: 7TB of satellite imagery](https://university.eugo.io/lesson/scaling-a-real-workload/the-workload)
- [Profile before you parallelize](https://university.eugo.io/lesson/scaling-a-real-workload/profiling-first)

### Distributing it

- [Partitioning 192,000 files](https://university.eugo.io/lesson/scaling-a-real-workload/partitioning)
- [When I/O is the bottleneck, not compute](https://university.eugo.io/lesson/scaling-a-real-workload/io-bound-vs-cpu-bound)

### Results

- [What we measured, and what surprised us](https://university.eugo.io/lesson/scaling-a-real-workload/what-we-measured)
- Cost accounting for the run (coming soon)

## Related reading

- [Cluster sizer](/interactive)