Skip to main content

Load-Shedding Policy

"When you can't serve everyone well, protect the requests that matter most."

Key Insight​

Under overload, load shedding deliberately refuses some requests early — using priority-aware admission control — so the most important traffic still meets its SLO. This project verifies that high-priority requests stay fast even at 2× overload.

Why This Matters​

When demand exceeds capacity, trying to serve everyone degrades the experience for everyone. Dropping low-priority work quickly — a fast rejection beats a slow timeout — keeps the requests that matter within their targets.