Unverified60% confidenceFactExact time

The token bucket algorithm allows instantaneous bursts (up to b requests passing immediately), but the long-term average rate does not exceed the fill rate r

2

Sources

60%

Confidence

Long-term

Relevance

8/14/2026

First Seen

Sources

Aquifer开源：为突发GPU推理负载削峰填谷的流量平滑方案

Related Claims

UnverifiedKeeping the model resident works by keeping the model in a resident process for reuse, avoiding the several-second to tens-of-seconds overhead of repeatedly loading multi-GB weights, and can push the response latency of subsequent requests down to the millisecond level.69% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/744917

API

curl https://kongchang.com/api/v1/knowledge/claims/744917

MCP

get_claim(id=744917)