Two Buckets One Burst

The same arrival pattern is sent through two limiters, both configured for 10 requests per second.

Token bucket: capacity 10 tokens, refilled at 10 tokens per second, one token per request, starts full. A request with no token available is rejected immediately.

Leaky bucket as a queue: a FIFO of depth 10 drained at exactly one request every 100 ms, the first drained at $t = 0$ . An arrival that finds the queue full is rejected.

25 requests arrive in a single instant at $t = 0$ , then nothing.

For each limiter, how many are admitted, how many are rejected, and when does the last admitted request reach the backend?

The token bucket admits 20, since the full bucket covers 10 and one second of refill covers 10 more, and the remaining 5 are rejected

Both admit 10 and reject 15, but the token bucket forwards all 10 to the backend at $t = 0$ while the leaky bucket releases one every 100 ms, so its last one arrives at about $t = 0.9$ s

The token bucket admits 10 and the leaky bucket admits all 25, because the leaky bucket queues excess arrivals instead of rejecting them

Both admit 10 and both hand the backend 10 requests per second, so the two are equivalent and differ only in implementation