Connection pooling in FastAPI - Performance & Optimization

Performance: Connection pooling

HIGH IMPACT

Connection pooling affects how quickly the server can handle database requests by reusing open connections instead of opening new ones each time.

Managing database connections efficiently in FastAPI

FastAPI

from asyncpg import create_pool

pool = await create_pool(dsn)

async def get_data():
    async with pool.acquire() as conn:
        result = await conn.fetch('SELECT * FROM table')
    return result

Reuses existing connections from the pool, avoiding repeated connection setup costs.

📈 Performance GainReduces request latency by 50-200ms, improves INP and server throughput

Managing database connections efficiently in FastAPI

FastAPI

async def get_data():
    conn = await asyncpg.connect(dsn)
    result = await conn.fetch('SELECT * FROM table')
    await conn.close()
    return result

Opening and closing a new database connection for every request adds significant latency and CPU overhead.

📉 Performance CostBlocks request handling for 50-200ms per connection setup, increasing INP and reducing throughput

Performance Comparison

Pattern	DOM Operations	Reflows	Paint Cost	Verdict
No connection pooling (new connection per request)	N/A	N/A	N/A	[X] Bad
Using connection pooling with asyncpg in FastAPI	N/A	N/A	N/A	[OK] Good

Rendering Pipeline

Connection pooling reduces backend wait times for database responses, speeding up the server's response generation before sending HTML or JSON to the browser.

→Server Processing

→Network Response

⚠️ BottleneckDatabase connection setup time

Core Web Vital Affected

INP

Connection pooling affects how quickly the server can handle database requests by reusing open connections instead of opening new ones each time.

Optimization Tips

1Always use connection pooling to reuse database connections in FastAPI.

2Avoid opening and closing connections per request to reduce latency.

3Monitor backend response times to verify pooling effectiveness.

Performance Quiz - 3 Questions

What is the main performance benefit of using connection pooling in FastAPI?

A Blocks rendering until all connections close

B Reduces latency by reusing open database connections

C Increases the number of database connections opened

D Improves CSS selector performance

DevTools: Network

How to check: Open DevTools Network panel, observe time to first byte (TTFB) for API calls to backend endpoints.

What to look for: Lower TTFB indicates faster backend response, showing effective connection pooling.

Practice

1. What is the main benefit of using connection pooling in FastAPI applications?

easy

A. It caches API responses for faster delivery

B. It reuses database connections to improve performance

C. It automatically creates database tables

D. It encrypts data sent to the database

Solution

1. Step 1: Understand connection pooling purpose

   Connection pooling keeps database connections open and reuses them instead of opening new ones each time.

2. Step 2: Identify the benefit in FastAPI context

   This reuse reduces the time spent opening connections, speeding up database access in FastAPI apps.

3. Final Answer:

   It reuses database connections to improve performance -> Option B

4. Quick Check:

   Connection pooling = reuse connections = better speed [OK]

Hint: Pooling means reusing connections, not creating or encrypting [OK]

Common Mistakes:

• Confusing pooling with encryption
• Thinking pooling creates tables automatically
• Mixing pooling with API response caching

2. Which of the following is the correct way to set the maximum pool size using SQLAlchemy in FastAPI?

easy

A. engine = create_engine(DB_URL, pool_size=10)

B. engine = create_engine(DB_URL, max_connections=10)

C. engine = create_engine(DB_URL, max_pool=10)

D. engine = create_engine(DB_URL, connection_limit=10)

Solution

1. Step 1: Recall SQLAlchemy pool size parameter

   The correct parameter to set max pool size is pool_size.

2. Step 2: Match parameter with options

   Only engine = create_engine(DB_URL, pool_size=10) uses pool_size=10, which is valid syntax.

3. Final Answer:

   engine = create_engine(DB_URL, pool_size=10) -> Option A

4. Quick Check:

   pool_size sets max connections in SQLAlchemy [OK]

Hint: Use pool_size to set max connections in create_engine [OK]

Common Mistakes:

• Using incorrect parameter names like max_connections
• Confusing pool_size with connection_limit
• Trying to set max_pool which is invalid

3. Given this FastAPI code snippet using SQLAlchemy, what will be the output when the app handles multiple requests?

from sqlalchemy import create_engine
engine = create_engine('sqlite:///test.db', pool_size=5, max_overflow=0)

# Each request uses engine.connect()

medium

A. Only one connection is used for all requests

B. Each request creates a new connection ignoring pool size

C. Up to 5 connections are reused; new requests wait if all are busy

D. The app crashes due to too many connections

Solution

1. Step 1: Understand pool_size effect

   Setting pool_size=5 means the engine keeps up to 5 connections open for reuse.

2. Step 2: Behavior on multiple requests

   When more than 5 requests come, new ones wait until a connection is free; no new connections beyond 5 are created.

3. Final Answer:

   Up to 5 connections are reused; new requests wait if all are busy -> Option C

4. Quick Check:

   pool_size=5 means max 5 reusable connections [OK]

Hint: pool_size limits max open connections reused [OK]

Common Mistakes:

• Assuming each request creates a new connection
• Thinking only one connection is shared
• Believing the app crashes when pool is full

4. You wrote this FastAPI code with SQLAlchemy connection pooling:

engine = create_engine('postgresql://user:pass@localhost/db', pool_size=5, max_overflow=0)
connection = engine.connect()
# forgot to close connection after use

What problem will this cause?

medium

A. No problem; connections close automatically

B. The database will reject all connections immediately

C. The app will automatically close connections after timeout

D. Connections will be exhausted causing new requests to hang

Solution

1. Step 1: Identify missing connection close

   Not closing connections means they stay checked out and unavailable for reuse.

2. Step 2: Effect on connection pool

   Since pool_size=5, after 5 connections are checked out and not closed, no free connections remain, causing new requests to wait indefinitely.

3. Final Answer:

   Connections will be exhausted causing new requests to hang -> Option D

4. Quick Check:

   Not closing connections exhausts pool causing hangs [OK]

Hint: Always close connections to free pool slots [OK]

Common Mistakes:

• Assuming connections close automatically without code
• Thinking database rejects connections immediately
• Believing app auto-closes connections after timeout

5. You want to optimize a FastAPI app using SQLAlchemy with a PostgreSQL database. The app has many short requests. Which pooling configuration best balances performance and resource use?

Options:
A) pool_size=20, max_overflow=10
B) pool_size=1, max_overflow=50
C) pool_size=5, max_overflow=0
D) pool_size=0, max_overflow=0

hard

A. pool_size=5, max_overflow=0

B. pool_size=1, max_overflow=50

C. pool_size=20, max_overflow=10

D. pool_size=0, max_overflow=0

Solution

1. Step 1: Understand pool_size and max_overflow roles

   pool_size sets fixed connections; max_overflow allows extra temporary connections beyond pool_size.

2. Step 2: Analyze options for many short requests

   Too large pool_size wastes resources; too small with high overflow risks overhead. pool_size=5, max_overflow=0 with moderate pool_size=5 and no overflow balances reuse and resource use well.

3. Final Answer:

   pool_size=5, max_overflow=0 -> Option A

4. Quick Check:

   Moderate pool_size with no overflow balances performance and resources [OK]

Hint: Moderate pool_size with zero overflow balances well [OK]

Common Mistakes:

• Choosing very high pool_size wasting resources
• Using zero pool_size disables pooling
• Setting high max_overflow causes many temporary connections