Useful Scripts for Testing

This folder contains practical scripts for load testing the ingestion and processing flow, plus comparing gzip vs plain request performance.

Scripts

send-gzip-events.ts

Sends gzip-compressed event batches to POST /api/v1/events.

• Sends bursts of REQUESTS_PER_BURST requests in parallel.
• Each request contains EVENTS_PER_REQUEST events.
• Starts a new burst every INTERVAL_SECONDS seconds.
• Runs continuously until you press Ctrl+C (first Ctrl+C stops gracefully after the current burst, second Ctrl+C exits immediately).
• Payloads are generated to satisfy the ingestion validator constraints.

Run:

bun scripts/send-gzip-events.ts

benchmark-gzip-vs-plain.ts

Benchmarks plain JSON requests against gzip-compressed JSON requests for the same event workload.

• Runs REQUESTS_PER_MODE rounds.
• Sends one plain and one gzip request per round.
• Alternates order to reduce measurement bias.
• Reports reliability, bandwidth, and latency (avg and p95).

Run:

bun scripts/benchmark-gzip-vs-plain.ts

Example Benchmark Run

--- Benchmark Summary ---
Endpoint: http://localhost:3000/api/v1/events
Requests per mode: 1000
Events per request: 100

Reliability:
plain: success=1000 failed=0
gzip:  success=1000 failed=0

Bandwidth:
plain sent bytes: 29.51 MB (30940393 B)
gzip sent bytes: 1.76 MB (1841165 B)
bandwidth saved with gzip: 94.05%

Latency (202 responses only):
plain avg=21.05ms p95=24.60ms
gzip  avg=21.06ms p95=24.52ms
avg latency gain with gzip: -0.02% (-0.00ms)
p95 latency gain with gzip: 0.35% (0.09ms)

Given the nature of our request body and the repetitive JSON structure/keys, we managed to reach an astonishing ~94% bandwidth reduction with gzip (this benchmark used a fixed payload size of 100 events per request; in a real setup where events are flushed in time windows, for example every 10 seconds, batch sizes will vary, so the 94% figure will likely decrease somewhat, but the improvement should still remain substantial).