عرض عملي: الخيوط الأساسية مقابل الافتراضية تحت الحِمل

خطوة بخطوة

1. Start Shop with 200 platform request threads; the endpoint calls a supplier with 500 ms latency and one DB query.
2. Run a load test with 1,000 concurrent clients; record throughput, p95 latency, and pool metrics.
3. Enable virtual threads, restart, and repeat.
4. Observe higher throughput, then database pool saturation (connections pending).
5. Planned mistake: also raise the DB pool to 1,000 "to fix it" and hit database limits. Fix: keep a sized pool or a Semaphore guard; accept queueing in the right place.
6. Summarize: the bottleneck moved from threads to the database.