Slide 13 — Live Production Metrics

Live Production Metrics — from EC2 & RDS Logs

Real Performance Data from Production Deployment

Metrics sourced from journalctl, RDS PostgreSQL, and systemd — EC2 ap-south-1 · Live since Feb 26, 2026 · 10 logged queries

Real Response Times — 10 Logged Queries (from journalctl)

1.2s

1.3s

1.4s

1.5s

1.9s

2.1s

2.6s

3.8s

5.9s

Q10

Metric (from logs)	Observed	Target	Status
⚡ Fastest Response	1.2s	< 3s	✅ Pass
📊 Median Response	1.7s	< 5s	✅ Pass
📈 Average Response	2.3s	< 5s	✅ Pass
🐢 P95 (Slowest)	5.9s	< 10s	⚡ OK
🩺 Crop Disease Query	3.8s	< 5s	✅ Pass
💬 General Query	1.2–2.6s	< 5s	✅ Pass
🧠 Memory (Streamlit)	287 MB	< 512MB	✅ Pass
🤖 Memory (Telegram)	178 MB	< 256MB	✅ Pass
💾 Disk Usage	4.7 GB / 7.6	< 80%	⚡ 63%
🔐 SSL Certificate	Let's Encrypt	HTTPS	✅ Active

RDS PostgreSQL — Live Database

Profiles

Chat History

Memories

Pwd Resets

7 tables · 8.7 KB · 4 unique users · SSL connection

🔄 LLM Failover — Real Events from Logs

BLOCKED

Gemini 2.0 Flash → hard-blocked

Feb 27 20:25

BLOCKED

Gemini 2.0 Flash-Lite → hard-blocked

Feb 27 20:25

SWITCH

Auto-switched → Groq Llama 3.3 70B

< 1s

BLOCKED

Gemini both models blocked (×3 events)

Feb 28

RESULT

Zero downtime — Groq auto-fallback success

100%

2.3s

Avg Response Time

From 10 logged queries
Median: 1.7s · P95: 5.9s

42h

Server Uptime

Since Feb 26, 20:05 UTC
Zero crash — all restarts = deploys

3/3

Failover Success

Gemini blocked → Groq auto-switch
3 real events logged

🖥️ Live Infrastructure

Instance

EC2 t2.micro (Mumbai)

CPU

2 vCPU Xeon 2.50GHz

RAM

914 MB (67% used)

Ubuntu 22.04 LTS

Proxy

Nginx + HSTS + SSL

Database

RDS PostgreSQL

LLM Active

Groq Llama 3.3 70B

Services

Streamlit + Telegram Bot

📋 Data Source

journalctl · psycopg2 · systemd
All metrics from live production