What really happens when 10,000 users hit your Spring Boot API simultaneously? Does Spring Boot create 10,000 threads? Does Tomcat queue requests? And can Java Virtual Threads genuinely change backend scalability?

If you’ve worked on Spring Boot systems in production, you’ve likely tuned thread pools, investigated latency spikes, or stared at CPU graphs wondering why throughput collapsed under load.

Understanding concurrency is no longer optional for backend engineers.

With Java 21 introducing Virtual Threads, the concurrency conversation has become even more interesting.

In this article, we’ll walk through:

Spring Boot request lifecycle
Tomcat worker threads
Platform Threads vs Virtual Threads
Concurrent execution behavior
Race conditions and synchronization fixes
Async execution patterns
Load testing concurrency
Production best practices

This write-up expands on Spring Boot concurrency concepts including thread handling, worker pools, and virtual threading models.

1. Understanding Spring Boot Request Lifecycle

Before discussing concurrency, we need to understand how a request flows through Spring Boot.

When an HTTP request arrives, Spring Boot does not magically process it by itself.

There is an embedded web server involved.

By default:

Spring Boot + MVC → Embedded Tomcat
Spring Boot + Reactive → Netty

For traditional MVC applications, Tomcat owns request execution.

Request Flow

Client Request
      │
      ▼
Tomcat Connector
      │
      ▼
Worker Thread Allocation
      │
      ▼
DispatcherServlet
      │
      ▼
Controller
      │
      ▼
Service Layer
      │
      ▼
Database / External API
      │
      ▼
HTTP Response

Here’s what actually happens:

Step 1 — Client Sends Request

A browser, mobile app, or downstream service calls:

GET /orders/123

Step 2 — Tomcat Accepts Connection

Tomcat listens on port 8080.

Incoming connections are accepted by Tomcat connectors.

Step 3 — Worker Thread Assignment

Tomcat picks a thread from its worker pool.

That thread becomes responsible for:

request parsing
controller invocation
business logic execution
database access
response writing

One request typically maps to one worker thread.

Step 4 — Spring MVC Processing

Tomcat invokes:

DispatcherServlet

DispatcherServlet routes request execution.

DispatcherServlet
      │
      ├── Handler Mapping
      ├── Controller
      ├── Service
      └── Response Serialization

Step 5 — Response Returned

Thread finishes work.

Tomcat returns thread to the pool.

Thread reuse begins.

2. Tomcat Worker Threads Deep Dive

Many developers think:

“Spring Boot handles concurrency.”

Technically true.

But Tomcat’s worker pool drives concurrency in Spring MVC systems.

Default Thread Pool Behavior

Tomcat manages:

worker threads
request queue
active connections

Common properties:

server:
  tomcat:
    threads:
      max: 200
      min-spare: 10
    accept-count: 100

What Do These Settings Mean?

max Threads

server.tomcat.threads.max=200

Maximum worker threads available.

If 300 requests arrive:

200 execute
remaining requests wait

min-spare Threads

Minimum idle threads maintained.

Helps reduce thread creation overhead.

accept-count

Overflow waiting queue.

Example:

200 threads busy
100 queued
new request arrives

Result:

connection refused

Restaurant Analogy

Think of Tomcat like a restaurant.

Customers = HTTP Requests
Waiters   = Worker Threads
Kitchen   = Business Logic
Queue     = accept-count

Scenario:

5 Waiters Available
20 Customers Arrive

Execution:

5 served immediately
10 waiting
5 rejected

This is essentially how overloaded APIs behave.

3. Concurrent Thread Execution Example (Java 8 Platform Threads)

Let’s see traditional concurrency.

Fixed Thread Pool Example

ExecutorService executor =
        Executors.newFixedThreadPool(3);

for(int i=1;i<=10;i++) {

    int task=i;

    executor.submit(() -> {

        System.out.println(
                Thread.currentThread().getName()
                +" processing "+task);

        Thread.sleep(2000);

        return null;
    });
}

What Happens Internally?

Only 3 platform threads exist.

Execution timeline:

Thread-1 → Task1
Thread-2 → Task2
Thread-3 → Task3

Task4 waits
Task5 waits
Task6 waits

After completion:

Thread-1 reused → Task4
Thread-2 reused → Task5

This is concurrent execution with bounded resources.

Sample Output

pool-1-thread-1 processing 1
pool-1-thread-2 processing 2
pool-1-thread-3 processing 3

pool-1-thread-1 processing 4
pool-1-thread-2 processing 5

Notice:

Threads are reused.

Not recreated.

4. The Cost of Platform Threads

Traditional Java threads are OS-backed threads.

Creating them is not free.

Each thread consumes:

native memory
stack allocation
OS scheduling cost
context switching overhead

Imagine:

50,000 concurrent requests

Creating 50,000 platform threads is usually a terrible idea.

Why?

Memory explosion.

Scheduler pressure.

CPU degradation.

Blocking I/O Problem

Suppose controller logic calls database.

@GetMapping("/users")
public List<User> users() {

    return repository.findAll();
}

During DB call:

Thread = BLOCKED

Worker thread waits doing nothing.

Yet memory is still consumed.

This becomes expensive at scale.

5. Enter Virtual Threads (Java 21)

Java 21 introduced Virtual Threads through Project Loom.

This changes concurrency economics.

Traditional Model

1 Request
   │
   ▼
1 Platform Thread
   │
   ▼
OS Scheduling

Virtual Thread Model

1 Request
   │
   ▼
1 Virtual Thread
   │
   ▼
JVM Scheduler
   │
   ▼
Small Platform Thread Pool

Virtual Threads are:

lightweight
JVM managed
cheap to create
optimized for blocking workloads

Massive Concurrency Example

try(var executor =
        Executors.newVirtualThreadPerTaskExecutor()) {

    for(int i=1;i<=100000;i++) {

        int task=i;

        executor.submit(() -> {

            Thread.sleep(1000);

            return task;
        });
    }
}

Traditional platform threads?

Probably impossible or highly unstable.

Virtual threads?

Designed for this model.

6. Platform Threads vs Virtual Threads

Feature	Platform Threads	Virtual Threads
Ownership	OS	JVM
Creation Cost	High	Low
Memory Usage	Higher	Lower
Blocking I/O	Expensive	Efficient
Large Scale Concurrency	Limited	Excellent
Existing Code Compatibility	Yes	Yes

Important Reality Check

Virtual Threads do not magically solve everything.

They help primarily when workloads are:

✅ I/O bound
✅ Database heavy
✅ API integration heavy

They do not automatically improve:

❌ CPU-bound computation
❌ inefficient SQL
❌ poor application architecture

7. Enabling Virtual Threads in Spring Boot 3.x

Spring Boot makes adoption surprisingly simple.

application.properties

spring.threads.virtual.enabled=true

That’s it.

Spring Boot begins using Virtual Threads where supported.

You can also create explicit executors.

@Bean
Executor executor() {

    return Executors
            .newVirtualThreadPerTaskExecutor();
}

8. Race Conditions: The Hidden Concurrency Bug

Concurrency is not just about speed.

It’s about correctness.

Consider:

@Service
public class CounterService {

    private int counter=0;

    public int increment() {

        counter++;

        return counter;
    }
}

Looks harmless.

Under concurrency?

Dangerous.

What Actually Happens

Suppose:

counter=20

Two requests execute simultaneously.

Thread-1

Reads:

Thread-2

Reads:

Thread-1

Writes:

Thread-2

Writes:

Expected:

Actual:

Classic race condition.

9. Fixing Race Conditions

Option 1 — AtomicInteger

Best lightweight fix.

private AtomicInteger counter =
        new AtomicInteger();

public int increment() {

    return counter.incrementAndGet();
}

Atomic operations guarantee thread safety.

Option 2 — synchronized

public synchronized int increment() {

    counter++;

    return counter;
}

Simple.

Reliable.

But can reduce throughput.

Option 3 — ReentrantLock

More control.

private Lock lock =
        new ReentrantLock();

public int increment() {

    lock.lock();

    try {

        counter++;

        return counter;

    } finally {

        lock.unlock();
    }
}

Useful for advanced concurrency control.

10. Async Execution in Spring Boot

Not every task should block request threads.

Examples:

email sending
PDF generation
notifications
report processing

Java 8 — @Async

Enable async.

@EnableAsync

Create async service.

@Async
public CompletableFuture<String> process() {

    Thread.sleep(3000);

    return CompletableFuture
            .completedFuture("done");
}

Controller returns quickly.

Heavy work continues separately.

Custom Executor

@Bean
Executor taskExecutor() {

    ThreadPoolTaskExecutor executor =
            new ThreadPoolTaskExecutor();

    executor.setCorePoolSize(10);
    executor.setMaxPoolSize(50);

    executor.initialize();

    return executor;
}

Production systems should almost always configure executors explicitly.

Java 21 Virtual Thread Async

@Bean
Executor executor() {

    return Executors
            .newVirtualThreadPerTaskExecutor();
}

Simpler scalability story.

Less thread pool tuning.

11. Load Testing Spring Boot Concurrency

Theory is useful.

Measurement matters more.

ApacheBench Example

ab -n 10000 -c 200 \
http://localhost:8080/orders

Parameters:

-n = total requests
-c = concurrency level

Example Test

Total Requests: 10,000
Concurrency: 200

Observe:

response time
throughput
active threads
CPU usage
GC activity

Common Failure Pattern

Under heavy load:

CPU low
Latency high
Requests waiting

Cause?

Usually blocked worker threads.

Not CPU exhaustion.

Virtual Threads can dramatically improve this scenario when waiting dominates execution time.

12. Production Lessons Learned

After enough production incidents, some patterns become obvious.

Lesson 1 — Bigger Thread Pools Are Not Always Better

Wrong instinct:

Performance issue?
Increase max threads to 5000.

Usually bad idea.

Symptoms:

memory pressure
context switching
degraded throughput

Lesson 2 — Database Pools Become Bottlenecks

Your application may support:

500 concurrent threads

But HikariCP:

maxPoolSize=20

Now 480 threads wait.

Concurrency shifted.

Problem not solved.

Lesson 3 — Blocking External APIs Matter

Third-party APIs frequently dominate latency.

Example:

App logic → 20ms
External API → 2.5 sec

Threads spend most of life waiting.

Ideal use case for Virtual Threads.

Lesson 4 — Measure Before Optimizing

Always collect:

thread dumps
heap metrics
CPU profiling
latency distribution
database timings

Performance tuning without metrics becomes guesswork.

13. When Should You Use What?

Use Platform Threads When

You have:

moderate traffic
mature Java 8 stack
CPU-heavy workloads
predictable thread requirements

Use Virtual Threads When

You have:

high concurrent traffic
blocking I/O
REST integrations
database waiting
Java 21 adoption path

Consider Reactive When

You need:

extreme scalability
event-driven streaming
fully non-blocking architecture

But remember:

Reactive introduces complexity.

Virtual Threads preserve familiar programming models.

Final Thoughts

Concurrency in Spring Boot is not just about adding more threads.

It’s about understanding:

request lifecycle
Tomcat worker behavior
blocking vs non-blocking execution
synchronization correctness
resource bottlenecks

Platform Threads served Java extremely well for years.

Virtual Threads do not replace engineering discipline.

But they dramatically improve how we think about scalable, blocking applications.

The next time someone says:

“Spring Boot handles concurrency automatically.”

You’ll know what’s actually happening underneath the hood.

1. Understanding Spring Boot Request Lifecycle

Request Flow

Step 1 — Client Sends Request

Step 2 — Tomcat Accepts Connection

Step 3 — Worker Thread Assignment

Step 4 — Spring MVC Processing

Step 5 — Response Returned

2. Tomcat Worker Threads Deep Dive

Default Thread Pool Behavior

What Do These Settings Mean?

max Threads

min-spare Threads

accept-count

Restaurant Analogy

3. Concurrent Thread Execution Example (Java 8 Platform Threads)

Fixed Thread Pool Example

What Happens Internally?

Sample Output

4. The Cost of Platform Threads

Blocking I/O Problem

5. Enter Virtual Threads (Java 21)

Traditional Model

Virtual Thread Model

Massive Concurrency Example

6. Platform Threads vs Virtual Threads

Important Reality Check

7. Enabling Virtual Threads in Spring Boot 3.x

application.properties

8. Race Conditions: The Hidden Concurrency Bug

What Actually Happens

Thread-1

Thread-2

Thread-1

Thread-2

9. Fixing Race Conditions

Option 1 — AtomicInteger

Option 2 — synchronized

Option 3 — ReentrantLock

10. Async Execution in Spring Boot

Java 8 — @Async

Custom Executor

Java 21 Virtual Thread Async

11. Load Testing Spring Boot Concurrency

ApacheBench Example

Example Test

Common Failure Pattern

12. Production Lessons Learned

Lesson 1 — Bigger Thread Pools Are Not Always Better

Lesson 2 — Database Pools Become Bottlenecks

Lesson 3 — Blocking External APIs Matter

Lesson 4 — Measure Before Optimizing

13. When Should You Use What?

Use Platform Threads When

Use Virtual Threads When

Consider Reactive When

Final Thoughts

Leave a Reply Cancel reply

Java – Design Patterns

Java 8 to 21 journey & Graal VM

Java – The basics