Cloud computing delivers storage, processing power, software, and managed services over the internet instead of requiring locally maintained hardware. A cloud provider operates remote data centers, while users obtain resources on demand and scale workloads up or down. This helps when growing data overwhelms a personal computer, without immediately buying more hardware. section 7.3.1 section 7.3
Cloud deployments can progress from a familiar local workflow to distributed computation:
- Run existing analytics on a cloud virtual machine, an isolated computer-like environment with its own operating system. This usually requires no code changes but offers only the capacity of one VM. section 7.3.2
- Configure additional resources or use managed data-processing services and serverless options such as Google Cloud Functions or AWS Lambda when a VM is insufficient. section 7.3.2
- Use cluster computing to divide work among interconnected nodes that process subtasks in parallel; Apache Hadoop, Apache Spark, and MPI provide established frameworks. section 7.3.1
Moving a calculation to the cloud
A practical prompt is: “How can I deploy my calculation to the cloud? Provide me with a walkthrough.” An AI assistant can provide instructions for a cloud VM, but provider interfaces and steps change, and an underspecified prompt may produce only a single-VM solution. A distance calculation between independent pairs of points is a better parallelization candidate because each pair can be processed independently. section 7.3.2 section 7.3.1
