Capacity Planner

Download SKILL.mdGet Murage How skills work

Open source. macOS, Windows and Linux.

At a glance

What it is

Capacity planning expertise covering load testing methodologies, autoscaling policies, resource forecasting, performance budgets, cost-capacity curves, bottleneck identification, queue theory basics, and building capacity models for production systems.

A skill is a written procedure a bot loads when a job calls for it. This one is a single file, SKILL.md, and the whole file is on this page.

File: SKILL.md
Length: 504 lines2,494 words, about 12 min read
Category: DevOps and cloudcloud infrastructure
Level: Intermediate
License

More DevOps and cloud skills.

The full skill.

The whole file, exactly as it downloads. Read it before you rely on it.

SKILL.md

504 lines · 2,494 words

Download

Capacity Planner

You are an expert capacity planner who ensures systems have enough resources to handle current load and predictable growth, without over-provisioning and wasting money. Capacity planning is the discipline that answers: "How much do we need, when will we need more, and how much will it cost?" Get it wrong on the low side and you have outages. Get it wrong on the high side and you waste thousands in unused infrastructure.

Capacity Planning Process

The Four Questions

1. WHERE ARE WE NOW?
   - Current resource utilization (CPU, memory, disk, network, connections)
   - Current request rate, latency, error rate
   - Current cost per unit of work

2. WHERE ARE WE GOING?
   - Traffic growth projections (historical trend + planned events)
   - Feature roadmap impact (new features that change load profile)
   - Seasonal patterns (Black Friday, tax season, school year)

3. WHEN WILL WE HIT LIMITS?
   - At current growth rate, when does each resource saturate?
   - Which resource saturates first? (the bottleneck)
   - What is the cost to push that limit?

4. WHAT SHOULD WE DO ABOUT IT?
   - Scale vertically (bigger instances)?
   - Scale horizontally (more instances)?
   - Optimize (reduce resource usage per request)?
   - Cache (reduce the number of requests that hit backend)?

Put this skill to work.

Questions.

What is the Capacity Planner skill?

How much does it cost?

Nothing. It ships with Murage, which is open source. Your bots run on the AI plan you already have.

Can I change it?

Yes. Open it in Settings → Skills to read, edit or duplicate it. Skill Guard scans it again after every edit.

Give your first job to Murage.

Download the app, connect the AI you already pay for, and tell your Chief of Staff what needs doing. Plan on about ten minutes from install to a working team.

Download Murage

No account needed · Runs on the AI plan you already pay for

1. WHERE ARE WE NOW? - Current resource utilization (CPU, memory, disk, network, connections) - Current request rate, latency, error rate - Current cost per unit of work 2. WHERE ARE WE GOING? - Traffic growth projections (historical trend + planned events) - Feature roadmap impact (new features that change load profile) - Seasonal patterns (Black Friday, tax season, school year) 3. WHEN WILL WE HIT LIMITS? - At current growth rate, when does each resource saturate? - Which resource saturates first? (the bottleneck) - What is the cost to push that limit? 4. WHAT SHOULD WE DO ABOUT IT? - Scale vertically (bigger instances)? - Scale horizontally (more instances)? - Optimize (reduce resource usage per request)? - Cache (reduce the number of requests that hit backend)?

Type	Purpose	How	Duration
Smoke test	Verify system works under minimal load	1-5 VUs, happy path	1-2 min
Load test	Validate performance at expected load	Expected VUs, realistic scenario	10-30 min
Stress test	Find the breaking point	Ramp up until failure	30-60 min
Soak test	Find memory leaks, resource exhaustion	Moderate load, extended time	2-12 hours
Spike test	Validate behavior under sudden traffic surge	Sudden jump to peak load	5-10 min
Breakpoint test	Find maximum capacity	Gradual increase until SLO breach	30-60 min

Metric	Good For	Watch Out For
CPU	Compute-bound services	IO-bound services show low CPU even when saturated
Memory	Memory-intensive workloads	Slow to reflect load changes (GC delays)
Request rate	Web servers, APIs	Does not account for request complexity
Queue depth	Worker/consumer services	Needs custom metrics pipeline
Latency	User-facing services	Lagging indicator (slow AFTER saturation)
Custom business metric	Specific workloads	Requires metrics instrumentation

Anti-Pattern	Problem	Fix
Scaling on CPU for IO-bound service	CPU stays low while service is saturated on DB connections	Scale on request latency or queue depth
Too aggressive scale-down	Pods removed, traffic spikes, pods added, repeat (thrashing)	Use stabilization window (5+ min for scale-down)
No minimum replicas	Scale to 0, cold start on first request	Set minReplicas >= 2 for availability
Scaling on average latency	One slow endpoint skews the metric	Scale on p95 or per-endpoint metrics
No max replicas	Runaway scaling costs a fortune	Always set maxReplicas with cost awareness

Strategy	Savings	Effort	Risk
Right-size instances	20-40%	Low	Low
Reserved/committed use	30-60%	Low	Medium (lock-in)
Spot/preemptible instances	60-80%	Medium	Medium (interruptions)
Autoscaling (scale down off-peak)	30-50%	Medium	Low
Caching (reduce backend calls)	40-70%	Medium	Low
Code optimization	10-50%	High	Low
Architecture change (e.g., serverless)	Varies	Very High	Medium

Capacity Planner

More DevOps and cloud skills.

The full skill.

Capacity Planner

Capacity Planning Process

The Four Questions

Put this skill to work.

Questions.

Give your first job to Murage.

Search murage.ai

Capacity Planner

More DevOps and cloud skills.

The full skill.

Capacity Planner

Capacity Planning Process

The Four Questions

Put this skill to work.

Questions.

Give your first job to Murage.

Create a bot and use this skill.

Ask Ember to create the bot

Open the bot’s skills

Find Capacity Planner and switch it on

Ask the bot for the job

About this skill.

Load Testing

Load Testing Types

k6 Load Test Examples

Interpreting Load Test Results

Autoscaling Policies

Horizontal Pod Autoscaler (Kubernetes)

Autoscaling Metric Selection

Autoscaling Anti-Patterns

Resource Forecasting

Capacity Model Template

Little's Law (Queue Theory Essential)

Performance Budgets

Defining Performance Budgets

Monitoring Budget Compliance

Cost-Capacity Curves

Modeling Cost per Request

Cost Optimization Strategies

Bottleneck Identification

The USE Method for Each Resource

Common Bottleneck Locations

Capacity Planning Checklist

When to Use

Output Format

Example

Edge Cases