Problems / LLM serving, medium
An experiment splits chats 50/50 between the current model (A) and a new version (B). B was deployed with only 60 concurrent streams; 40 chats arrive a second.
Unlock it once for 15 credits; everything inside it is free (runs, changes, Submit, results, hints). Sign in with Google to start; new accounts get free credits.
Success ≥ 99%.
Like a coding problem, a solution has to pass every case, not just the one it was built for. Submit in the lab runs all 6; 2 of them are hidden until then.
40 requests a second for 30 seconds: 100% request (reads).
What each component does in this design, and what calls it. The simulator mimics each one from these settings.
| Component | Its job here | Settings |
|---|---|---|
| Experiment splitModel router | Traffic enters here.Calls Variant A app, Variant B app.A router sends easy requests to small, cheap models and hard ones to larger models. | 100000 slots |
| Variant A appApp service | Experiment split forwards to it 50% of traffic.Calls Model A (current).Little's law: throughput ≈ concurrency ÷ latency. 200 slots at 50 ms serve about 4,000 req/s. | 3 replicas × 500 slots · timeout 30 s |
| Model A (current)LLM endpoint | Variant A app calls it.Latency has two parts: time to the first token, then a steady stream of tokens per second. | 400 streams |
| Variant B appApp service | Experiment split forwards to it 50% of traffic.Calls Model B (candidate).Little's law: throughput ≈ concurrency ÷ latency. 200 slots at 50 ms serve about 4,000 req/s. | 3 replicas × 500 slots · timeout 30 s |
| Model B (candidate)LLM endpoint | Variant B app calls it.Latency has two parts: time to the first token, then a steady stream of tokens per second. | 60 streams |