Thundering Herd & Retry Storm Simulator

Initial Outage Event

Simultaneous Timed-out Requests 10,000

Client Retry Policy

Backoff Algorithm

Max Retries per Client 3

Base Delay / Slot (ms) 1000 ms

Max Backoff Ceiling (ms) 16,000 ms

Backend Server Bounds

Server Max Capacity (RPS) 3,000

Normal Background Access (RPS) 1,000

Peak Total Load

11,000 RPS

366% of server capacity

Retry Storm Amplification

1.42x

4,200 retries executed

Dropped Requests

0.0% failure rate

System Recovery Time

6.8s

Queue fully cleared

Real-Time Request Traffic & Server Queue Timeline

            Timeline window: 20 seconds
          

🔍 System Observation & Mitigation Proof

Applying Exponential Backoff + Full Jitter distributes initial retries uniformly across time intervals, completely preventing synchronous secondary stampedes. Peak incoming load stays within server capacity, stabilizing queue depth and allowing normal background throughput to recover smoothly.