Virtual Threads vs Platform Threads

Comparison

	Platform thread	Virtual thread
Managed by	OS	JVM
Memory	~1 MB stack reserved	small, grows on the heap
Creation cost	expensive (system call)	cheap (object allocation)
How many	thousands	millions
Pool them?	yes (thread pools)	no — create one per task
Best for	CPU-bound work	I/O-bound, blocking code
Daemon	configurable	always daemon
ThreadLocal	fine	works, but millions of copies can use a lot of memory

Virtual Threads vs Reactive

Before Loom, high-concurrency Java used reactive libraries (WebFlux, RxJava): non-blocking but hard to read and debug. Virtual threads give similar scalability with plain, blocking, step-by-step code and normal stack traces.

// Reactive (WebFlux)
return userRepo.findById(id)
        .flatMap(user -> orderClient.ordersFor(user.id())
                .map(orders -> new Profile(user, orders)));

// Virtual threads: same scalability, ordinary code
User user = userRepo.findById(id).orElseThrow();
List<Order> orders = orderClient.ordersFor(user.id());
return new Profile(user, orders);