Automate benchmarks for various tiling scenarios

[ividito]

Description

Ref: #36

In order to provide performance metrics for our tiling service, we want to be able to automatically benchmark both local API runtimes and our hosted environments by simulating typical usage patterns. To start, these three patterns should be:

• Requesting several tiles from a collection containing a single COG (tiles selected pseudo-randomly, with option to include known 404s)
• Requesting several tiles from a mosaic containing multiple COGs
• Requesting a small set of tiles, multiple times (simulating a discovery, where a specific set of tiles are used to showcase relevant data)

When running these benchmarks on an AWS-hosted API, we will generate several metrics and logs that can be analyzed in Cloudwatch. To distinguish benchmark requests from typical user requests, a pre-determined correlation ID should be included with all requests.

AC

• Benchmark script is configurable (can easily modify target API and collection IDs to be requested)
• Benchmarks for 3 identified scenarios
• Benchmarks use a deterministic correlation-id to be used for AWS Cloudwatch analysis.
• Benchmarks are modular, keeping in mind potential load testing use cases in the future.