Async vs sync decision in FastAPI - Performance Comparison

Performance: Async vs sync decision

HIGH IMPACT

This affects how quickly the server can respond to requests and handle multiple users at once.

Handling multiple web requests that involve waiting for database or network responses

FastAPI

from fastapi import FastAPI
import asyncio
app = FastAPI()

@app.get('/async')
async def read_async():
    await asyncio.sleep(2)  # non-blocking wait
    return {'message': 'done'}

Async function releases the event loop during wait, allowing other requests to be processed simultaneously.

📈 Performance GainImproves concurrency and responsiveness, reducing request wait times under load

Handling multiple web requests that involve waiting for database or network responses

FastAPI

from fastapi import FastAPI
app = FastAPI()

@app.get('/sync')
def read_sync():
    import time
    time.sleep(2)  # blocking wait
    return {'message': 'done'}

The synchronous function blocks a worker thread during the sleep, limiting concurrent requests.

📉 Performance CostBlocks worker thread, causing limited concurrency and increased response time under load

Performance Comparison

Pattern	Concurrency	Event Loop Blocking	Response Delay	Verdict
Synchronous endpoint	Limited by thread pool size	Yes (blocks threads during I/O)	High under load	[X] Bad
Asynchronous endpoint	Multiple requests concurrently	Non-blocking event loop	Low even under load	[OK] Good

Rendering Pipeline

In FastAPI, async endpoints allow the server to handle other requests while waiting for I/O. Sync endpoints run in a thread pool and block worker threads during waits, which can limit concurrency under load.

→Request Handling

→Thread Pool

→Response Sending

⚠️ BottleneckWorker thread blocking in synchronous code

Core Web Vital Affected

INP

This affects how quickly the server can respond to requests and handle multiple users at once.

Optimization Tips

1Use async endpoints for I/O-bound operations to keep the server responsive.

2Avoid blocking calls like time.sleep or long computations in sync endpoints.

3Test under load to see if sync code causes request delays or timeouts.

Performance Quiz - 3 Questions

What is the main performance benefit of using async endpoints in FastAPI?

A They allow handling multiple requests without waiting for I/O operations to finish.

B They reduce the size of the server code.

C They make CPU-bound tasks faster.

D They automatically cache responses.

DevTools: Network and Performance panels

How to check: Use the Network panel to observe request timings and the Performance panel to record server response delays under load.

What to look for: Look for long blocking times and queued requests indicating sync blocking; async shows overlapping request handling.

Practice

1. In FastAPI, when should you choose an async function over a def synchronous function?

easy

A. When the task is CPU-heavy and requires intense calculations

B. When the task involves waiting for external resources like databases or APIs

C. When you want the function to run slower

D. When you want to block other tasks until this finishes

Solution

1. Step 1: Understand async purpose in FastAPI

   Async functions let your app wait for external tasks like database calls without stopping other work.

2. Step 2: Match task type to function type

   Waiting for external resources is I/O-bound, so async is best to avoid blocking.

3. Final Answer:

   When the task involves waiting for external resources like databases or APIs -> Option B

4. Quick Check:

   Async = I/O-bound tasks [OK]

Hint: Use async for waiting tasks, sync for CPU tasks [OK]

Common Mistakes:

• Choosing async for CPU-heavy tasks
• Thinking async makes code slower
• Using sync for I/O-bound tasks

2. Which of the following is the correct way to define an asynchronous route handler in FastAPI?

easy

A. async get_data def(): return {'msg': 'Hello'}

B. def async get_data(): return {'msg': 'Hello'}

C. def get_data() async: return {'msg': 'Hello'}

D. async def get_data(): return {'msg': 'Hello'}

Solution

1. Step 1: Recall Python async function syntax

   Async functions start with async def followed by the function name and parentheses.

2. Step 2: Check each option's syntax

   Only async def get_data(): return {'msg': 'Hello'} uses the correct async def syntax; others are invalid Python syntax.

3. Final Answer:

   async def get_data(): return {'msg': 'Hello'} -> Option D

4. Quick Check:

   Async function syntax = async def [OK]

Hint: Async functions start with 'async def' in Python [OK]

Common Mistakes:

• Placing async keyword after def
• Using 'def async' instead of 'async def'
• Incorrect order of keywords

3. Consider this FastAPI code snippet:

from fastapi import FastAPI
import time

app = FastAPI()

@app.get('/sync')
def sync_endpoint():
    time.sleep(2)
    return {'message': 'Done'}

@app.get('/async')
async def async_endpoint():
    await some_async_task()
    return {'message': 'Done'}

What is the main difference in behavior between /sync and /async endpoints?

medium

A. Both block other requests equally

B. /async blocks other requests; /sync allows concurrency

C. /sync blocks other requests during sleep; /async allows concurrent handling

D. Both allow full concurrency without blocking

Solution

1. Step 1: Understand sync endpoint behavior

   The sync_endpoint uses time.sleep(2), which blocks the thread and stops handling other requests during that time.

2. Step 2: Understand async endpoint behavior

   The async_endpoint uses await on an async task, allowing the server to handle other requests while waiting.

3. Final Answer:

   /sync blocks other requests during sleep; /async allows concurrent handling -> Option C

4. Quick Check:

   Sync blocks, async allows concurrency [OK]

Hint: Sync blocks during wait; async frees server to handle others [OK]

Common Mistakes:

• Assuming sync allows concurrency
• Thinking async blocks like sync
• Confusing await with blocking

4. You wrote this FastAPI route:

@app.get('/data')
async def get_data():
    result = time.sleep(1)
    return {'result': result}

What is the problem with this code?

medium

A. Using time.sleep inside async blocks the event loop, causing delays

B. Missing await keyword before time.sleep

C. Function should not be async if it returns a dictionary

D. No problem; code works fine

Solution

1. Step 1: Identify blocking call in async function

   time.sleep(1) is a blocking call that pauses the whole thread, including the async event loop.

2. Step 2: Understand async best practices

   In async functions, use non-blocking calls like await asyncio.sleep(1) to avoid blocking the event loop.

3. Final Answer:

   Using time.sleep inside async blocks the event loop, causing delays -> Option A

4. Quick Check:

   Blocking calls in async cause delays [OK]

Hint: Never use time.sleep in async; use await asyncio.sleep instead [OK]

Common Mistakes:

• Trying to await time.sleep
• Ignoring blocking effect of time.sleep
• Thinking async auto makes code non-blocking

5. You have a FastAPI app that fetches data from a slow external API and then processes it heavily with CPU calculations. How should you design your endpoint for best performance?

hard

A. Make the endpoint async to await the API call, then run CPU work in a separate thread or process

B. Make the endpoint fully async and run CPU work directly inside it

C. Make the endpoint synchronous and run both API call and CPU work sequentially

D. Use synchronous endpoint but call the API asynchronously

Solution

1. Step 1: Separate I/O-bound and CPU-bound tasks

   Fetching data from an external API is I/O-bound and benefits from async to avoid blocking.

2. Step 2: Handle CPU-heavy work properly

   CPU-heavy processing blocks the event loop if run directly in async code, so it should run in a separate thread or process.

3. Step 3: Combine async I/O with offloaded CPU work

   Use async endpoint to await API call, then offload CPU work to thread/process to keep responsiveness.

4. Final Answer:

   Make the endpoint async to await the API call, then run CPU work in a separate thread or process -> Option A

5. Quick Check:

   Async for I/O + thread/process for CPU [OK]

Hint: Async for waiting; thread/process for CPU work [OK]

Common Mistakes:

• Running CPU work directly in async code
• Making endpoint fully sync causing blocking
• Mixing async and sync calls incorrectly