Protect a fixed GPU fleet at launch

  • Survive
  • Agent 24

Difficulty 5 of 5

A chat app launches on 2 H100 GPUs (about 24 chats/s). Demand rises from 16 to 64 chats/s in 4 waves, and the last wave has a 10 s burst at 6 times the rate. Keep p99 under 7 s and errors under 15 percent for 90 percent of the run. Budget: 3,812 USD/month, plus 4 per wave.

Play level

1,900 USD/mo1,900 USD/moHTTP/2Users16 req/sAPI server1 vCPU · 1 GBLoad balanc…1 vCPU · 1 GBLLM server1 vCPU · 1 GBLLM server1 vCPU · 1 GB
1,900 USD/mo1,900 USD/moHTTP/2HTTP/2Users16 req/sAPI serverLoad balanc…LLM serverLLM server
The board at the start of the level
One server boardDrag or use arrow keys to change the load

280 req/s · CPU 70% · mean 22 ms

Goals and limits

Time within SLO
at least 90%
p99 latency
under 7 s
Error rate
under 15%
Monthly cost
under 3,812 USD/mo
Waves
4
Time within SLO
at least 90%
Parts you can add
Semantic cache, Redis cache, Rate limiter
Part limit
2
Hosts
2 × 1 vCPU · 80 GB
Traffic
16 req/s

Concepts

How players did

Player results load from the server.

All levels