When to use microservices (and when not to) - Scalability & System Analysis

Scalability Analysis - When to use microservices (and when not to)

Growth Table: Microservices Adoption at Different Scales

Users	System Complexity	Team Size	Deployment Frequency	Service Boundaries
100 users	Simple, monolith works well	Small (1-5)	Low to medium	Not needed
10,000 users	Growing complexity, some modularity	Medium (5-20)	Medium	Consider splitting key modules
1,000,000 users	High complexity, many features	Large (20+)	High, frequent releases	Microservices beneficial
100,000,000 users	Very high complexity, global scale	Very large, multiple orgs	Continuous delivery	Microservices essential for scale

First Bottleneck: When Microservices Are Needed

At small scale, the main bottleneck is development speed and coordination. A monolith is easier to manage.

As users and features grow, the codebase becomes large and hard to maintain. Teams slow down due to dependencies.

The first bottleneck is the development and deployment velocity caused by tightly coupled code and shared databases.

Microservices help by splitting the system into smaller, independent parts owned by small teams.

Scaling Solutions: When to Use Microservices

• Use microservices when:
  • System complexity is high with many features.
  • Multiple teams need to work independently.
  • Frequent, independent deployments are required.
  • Different parts of the system have different scaling needs.
  • Technology diversity is beneficial (different languages/databases per service).
• Do NOT use microservices when:
  • System is small or simple.
  • Team size is small and communication is easy.
  • Operational overhead of microservices is too high.
  • Latency between services would hurt user experience.
  • Strong consistency and transactions across services are critical and complex.

Back-of-Envelope Cost Analysis

At 1,000 users, a single server and database handle ~100-500 requests per second easily.

At 10,000 users, requests grow to ~1,000-5,000 QPS; a monolith with caching can still work.

At 1,000,000 users, requests can reach ~100,000 QPS; a single database struggles with this load.

Microservices allow splitting load across multiple databases and servers, reducing bottlenecks.

Operational costs increase due to multiple deployments, monitoring, and network overhead.

Interview Tip: Structuring Your Scalability Discussion

Start by describing the current system scale and team size.

Explain the pain points: slow deployments, tight coupling, scaling issues.

Identify the first bottleneck: development velocity or database load.

Propose microservices only if complexity and team size justify it.

Discuss trade-offs: operational overhead vs. independent scaling.

Self-Check Question

Your database handles 1000 QPS. Traffic grows 10x to 10,000 QPS. What do you do first?

Answer: Add read replicas and caching to reduce database load before splitting into microservices.

Key Result

Microservices are best used when system complexity and team size grow large enough that a monolith slows development and scaling; otherwise, they add unnecessary overhead.

Practice

1. Which scenario is best suited for using microservices architecture?

easy

A. A large, complex application requiring independent scaling of components

B. A simple, single-function app with a small user base

C. A small script running on a single machine

D. A static website with no backend logic

Solution

1. Step 1: Understand microservices purpose

   Microservices are designed for complex apps where parts can scale or update independently.

2. Step 2: Match scenario to microservices benefits

   A large app needing flexibility and scaling fits microservices well; small or simple apps do not.

3. Final Answer:

   A large, complex application requiring independent scaling of components -> Option A

4. Quick Check:

   Complex app = microservices [OK]

Hint: Use microservices only for complex, scalable apps [OK]

Common Mistakes:

• Choosing microservices for small or simple apps
• Ignoring team size and management overhead
• Assuming microservices always improve performance

2. Which of the following is a correct reason NOT to use microservices?

easy

A. The app requires frequent updates to parts

B. The application is very small and simple

C. The app needs to scale independently

D. The app has multiple teams working on different features

Solution

1. Step 1: Identify when microservices are unnecessary

   Microservices add complexity and overhead, so small simple apps don't benefit.

2. Step 2: Evaluate options

   Options A, B, and D are reasons to use microservices, not avoid them.

3. Final Answer:

   The application is very small and simple -> Option B

4. Quick Check:

   Small app = avoid microservices [OK]

Hint: Avoid microservices for small, simple apps [OK]

Common Mistakes:

• Confusing scaling needs as a reason to avoid microservices
• Ignoring complexity added by microservices
• Assuming microservices fit all team sizes

3. Consider a microservices app with 5 services. If each service requires 2 developers and the team has only 6 developers total, what is the likely outcome?

medium

A. The team can easily manage all services independently

B. The services will merge into a monolith automatically

C. The team will struggle due to insufficient resources for each service

D. The app will automatically scale without developer input

Solution

1. Step 1: Calculate developer needs

   5 services x 2 developers each = 10 developers needed.

2. Step 2: Compare with available team size

   Only 6 developers are available, which is less than 10, causing resource strain.

3. Final Answer:

   The team will struggle due to insufficient resources for each service -> Option C

4. Quick Check:

   Dev shortage = struggle managing microservices [OK]

Hint: Check if team size matches microservices needs [OK]

Common Mistakes:

• Assuming microservices scale developer needs automatically
• Ignoring team size constraints
• Thinking services merge automatically without effort

4. A team tries to convert a small monolithic app into microservices but faces deployment failures and communication errors. What is the most likely cause?

medium

A. Microservices do not support deployment automation

B. The app was too large for microservices

C. They used too many developers

D. They underestimated the complexity of managing microservices

Solution

1. Step 1: Analyze the problem context

   Small apps converted to microservices often face complexity in communication and deployment.

2. Step 2: Identify the cause

   Deployment failures and communication errors usually come from underestimating microservices management overhead.

3. Final Answer:

   They underestimated the complexity of managing microservices -> Option D

4. Quick Check:

   Underestimating complexity = deployment issues [OK]

Hint: Expect extra management work with microservices [OK]

Common Mistakes:

• Blaming microservices for deployment automation lack
• Assuming more developers cause deployment errors
• Thinking large apps cause these specific errors

5. A startup with a small team plans to build a new app. They want to decide between microservices and a monolithic design. Which approach should they choose and why?

hard

A. Start with a monolith to reduce complexity and switch later if needed

B. Start with microservices to prepare for future scaling immediately

C. Use microservices only if the app is a static website

D. Avoid both and build multiple separate apps

Solution

1. Step 1: Consider team size and app complexity

   A small team benefits from simpler monolithic design to reduce overhead and speed development.

2. Step 2: Plan for future growth

   Starting monolithic allows easier initial development; microservices can be adopted later if scaling is needed.

3. Final Answer:

   Start with a monolith to reduce complexity and switch later if needed -> Option A

4. Quick Check:

   Small team = start monolith [OK]

Hint: Small teams start monolith, scale to microservices later [OK]

Common Mistakes:

• Choosing microservices too early for small teams
• Confusing static websites with microservices use
• Ignoring future scalability planning