Quiz: Systems Thinking

← Back to quiz index

3 questions

L1 (1 questions)

1. What is a positive (reinforcing) feedback loop in infrastructure? Give an example.

Show answer

A positive feedback loop amplifies change rather than stabilizing the system. Example: a retry storm — Service A times out on B, retries double the load on B, B slows further, more timeouts cause more retries, load quadruples, B crashes, A's queues fill, A crashes too. The loop has no natural stopping point and amplifies until something breaks. Fix with circuit breakers, backoff, and rate limits.

L2 (2 questions)

1. Explain Little's Law (L = lambda * W) and why a 'small' latency increase can cause a total outage.

Show answer

L (concurrent requests) = lambda (arrival rate) * W (average latency). If your API handles 100 rps at 200ms, L = 20 concurrent requests. If latency increases to 2 seconds (database hiccup), L jumps to 200. Your connection pool of 50 is exhausted, requests queue, latency increases further, L grows more — a positive feedback loop. A 10x latency increase requires 10x capacity to handle the same throughput.

2. Why can adding retries to a failing service make the failure worse? What is the systems-aware alternative?

Show answer

If Service B is failing at 10% because it is overloaded, adding 3 retries from Service A increases load on B by up to 3x. B's failure rate rises from 10% to 40% or more, and with retries the effective load is 4x original — B crashes entirely. The fix looks correct in component thinking but is catastrophic in systems thinking. The alternative is a retry budget: only retry if total retry rate is below 10% of requests. Otherwise, fail fast.