Data Collection — Implementing System & Sensor Monitoring

🎯 Overview

In this post, we implement the data collection layer, which is the foundation of the entire monitoring system.

We collect two types of data:

• System metrics (CPU, Memory, Disk)
• Hardware sensor data (Temperature, Fan Speed, Power)

Due to platform limitations, we adopt a hybrid collection strategy.

---

🏗️ Data Collection Architecture

[System Metrics]
Telegraf (60s interval)
        ↓
InfluxDB

[Sensor Data]
HWiNFO → Python (3s interval)
        ↓
InfluxDB

---

📡 1. System Metrics Collection (Telegraf)

Why Telegraf?

• Lightweight agent
• Native InfluxDB integration
• Supports Windows

⚙️ Telegraf Configuration

[agent]
  interval = "60s"

[[inputs.cpu]]
  percpu = true
  totalcpu = true

[[inputs.mem]]

[[inputs.disk]]

[[outputs.influxdb_v2]]
  urls = ["http://localhost:8086"]
  token = "YOUR_TOKEN"
  organization = "YOUR_ORG"
  bucket = "iot"

---

🌡️ 2. Sensor Data Collection (HWiNFO + Python)

Why Not Telegraf?

Telegraf cannot collect:

• Temperature
• Fan Speed
• Power

Solution

Use:

• HWiNFO (Sensor Provider)
• Python (Collector)

⚙️ HWiNFO Setup

Steps:

1. Run HWiNFO in Sensors-only mode
2. Enable:
   • Shared Memory Support

---

🧠 3. Hybrid Strategy Explained

Data Type	Solution	Interval
CPU / Memory	Telegraf	60s
Temperature	Python	3s
Fan Speed	Python	3s

⚠️ Challenges

1. Windows Limitation
   • No native API for sensors
   • Requires external tool
2. Data Synchronization
   • Different intervals
   • Requires aggregation later
3. Performance
   • High-frequency writes → InfluxDB tuning needed

🎯 Key Takeaways

1. Use Telegraf for standard metrics
2. Use Python for custom sensor ingestion
3. Separate pipelines for flexibility

---

🚀 Next Step

In the next post, we will design the data storage schema, including:

InfluxDB measurement design MySQL aggregation tables