📊 Performance Benchmarking Report — firmflow.¶
Date: February 2026
Hardware Profile: 8-Core CPU, 16GB RAM (Standard mid-range server)
Concurrent Users: 50+
1. Executive Summary¶
Load testing was conducted using Artillery to simulate realistic traffic across the local client portal and the firm admin dashboard. The platform maintained high availability and sub-second response times under sustained load.
2. Test Results¶
Latency Summary¶
| Metric | Warm Up (5-20 req/s) | Sustained Load (20 req/s) |
|---|---|---|
| Response Time (p95) | 115ms | 185ms |
| Response Time (p99) | 240ms | 410ms |
| Success Rate | 100% | 99.8% |
Resource Utilization¶
- CPU Usage: Average 35% during sustained load. Peaks at 60% during heavy PDF processing/encryption.
- Memory Usage: Stable at 1.2 GB (Node.js heap) + 800 MB (PostgreSQL cache).
3. Bottleneck Analysis¶
- PDF Generation: Massive concurrent PDF merges (e.g., during bulk sign-off) can spike CPU. Recommended handling via queuing in enterprise deployments with 100+ concurrent staff.
- Database I/O: High-volume audit log writes benefit from SSD storage. p99 latency increased by 15% when using spinning disks (NAS).
4. Scaling Recommendations¶
- Scale Out: Use the
kubectl scaleinstructions in the Technical Manual for firm-wide deployment. - Queue Workers: In Phase 6 (Future), move document analysis to a separate background service to prevent UI blocking.
Report generated for Toraa Global Release v1.2.0.