Penjelasan Jenis-Jenis Metrik

Welcome to "Types of Metrics Explained"! In observability, metrics are crucial for understanding your system's health and performance.

But not all numbers are the same! Categorizing metrics helps us collect, store, and analyze them effectively.

We'll explore the four fundamental types: Gauges, Counters, Histograms, and Summaries.

A Gauge represents a single numerical value that can go up and down over time. Think of it like a car's speedometer or a thermometer.

Gauges are perfect for capturing the current state of a system at a specific moment.

• Use for: Current CPU usage, memory consumption, queue size, temperature.
• Nature: Point-in-time value.

Here's a simple Java example simulating a gauge tracking current CPU utilization. Notice how its value can change freely.

A Counter is a cumulative metric that only ever increases. It represents a total count of something that has occurred since the system started.

It can reset to zero only when the monitored system restarts.

• Use for: Total requests served, errors encountered, bytes sent, login attempts.
• Nature: Monotonically increasing total.

This Java example demonstrates a counter for total requests. Each 'request' simply increments the counter.

Histograms sample observations (like request durations or response sizes) and count them in configurable buckets.

They give you insight into the distribution of values, not just the average. This is vital for understanding latency and performance.

• Benefit: You can calculate percentiles (e.g., 99th percentile latency) on the server side.
• Use for: Request latency, response sizes, data transfer rates.

Summaries are similar to histograms but often pre-calculate configurable quantiles (like p99, p95, p50) on the client side.

Instead of sending raw data, the client library sends pre-computed statistics (sum, count, and quantiles) to the monitoring system.

• Benefit: Less data sent over the network, but less flexible for custom percentile calculations later.
• Use for: Latency measurements where specific percentiles are known to be needed.

Both Histograms and Summaries track distributions, but they differ in where calculations happen:

• Histograms: Send raw data (counts in buckets). Percentiles are calculated on the server. More flexible for ad-hoc analysis.
• Summaries: Calculate percentiles on the client and send pre-computed results. More resource-efficient if you know exactly which percentiles you need.

For most modern systems, Histograms are generally preferred due to their flexibility.

Choosing the right metric type is key for effective monitoring:

• Gauges: For current values that can go up/down (e.g., disk usage, active users).
• Counters: For cumulative totals that only increase (e.g., total errors, processed items).
• Histograms: For distributions of values where you need server-side percentile calculation and flexibility (e.g., request durations).
• Summaries: For distributions where client-side pre-calculated percentiles are sufficient and network efficiency is critical (less common than histograms now).

Great job! You've now learned about the four fundamental metric types:

• Gauges: For current, fluctuating values.
• Counters: For cumulative, ever-increasing totals.
• Histograms: For understanding the distribution of values and calculating percentiles server-side.
• Summaries: For pre-calculated percentiles client-side.

Understanding these types helps you choose the right tool for the right job, leading to more insightful monitoring and faster debugging!