Skip to main content
Version: 2.0.x

Basic Concurrency

ZIO is a highly concurrent framework, powered by fibers, which are lightweight virtual threads that achieve massive scalability compared to threads, augmented with resource-safe cancellation, which powers many features in ZIO.

This powerful concurrency model lets you do more with less, achieving highly-scalable, ultra low-latency applications that are globally efficient and resource-safe.

In this section, you will learn the basics of fibers, and become acquainted with some of the powerful high-level operators that are powered by fibers.


All effects in ZIO are executed by some fiber. If you did not create the fiber, then the fiber was created by some operation you are using (if the operation is concurrent or parallel), or by the ZIO runtime system.

Even if you only write "single-threaded" code, with no parallel or concurrent operations, there will be at least one fiber: the "main" fiber that executes your effect.

Like operating system-level threads, ZIO fibers have a well-defined lifecycle, defined by the effect they are executing.

Every fiber exits with failure or success, depending on whether the effect it is executing fails or succeeds.

Also like operating system threads, ZIO fibers have unique identities, stacks (including stack traces), local state, and a status (such as done, running, or suspended).

Compared to operating system threads, ZIO fibers:

  • Consume almost no memory
  • Have dynamic stacks that grow and shrink
  • Don't waste operating system threads with blocking operations
  • Can be safely interrupted at any point in time
  • Are strongly typed
  • Let you query them to discover their children
  • Will be garbage collected automatically if they are suspended and cannot be reactivated

These make fibers a superior choice for building modern applications.

Fibers are scheduled onto operating system threads by the ZIO runtime. Because fibers cooperatively yield to each other, ZIO fibers always execute concurrently, even when running in a single-threaded environment like JavaScript (or the JVM, when ZIO is configured with one work thread).

The Fiber Data Type

The Fiber data type in ZIO represents a "handle" on the execution of an effect. The Fiber data type is most similar to Scala's Future data type, which represents a "handle" on a running asynchronous operation.

The Fiber[E, A] data type in ZIO has two type parameters:

  • E Failure Type. The fiber may fail with a value of this type.
  • A Success Type. The fiber may succeed with a value of this type.

Fibers do not have an R type parameter, because fibers only execute effects that have already had their requirements provided to them.

Forking Effects

The most fundamental way of creating a fiber is to take an existing effect and fork it. Conceptually, forking an effect begins executing the effect on a new fiber, giving you a reference to the newly-created fiber.

The following code creates a single fiber using fork, which executes fib(100) independently of the main fiber:

def fib(n: Long): UIO[Long] = 
ZIO.suspendSucceed {
if (n <= 1) ZIO.succeed(n)
else fib(n - 1).zipWith(fib(n - 2))(_ + _)

val fib100Fiber: UIO[Fiber[Nothing, Long]] =
for {
fiber <- fib(100).fork
} yield fiber

Joining Fibers

One of the methods on Fiber is Fiber#join, which returns an effect. The effect returned by Fiber#join will succeed or fail as per the fiber:

for {
fiber <- ZIO.succeed("Hi!").fork
message <- fiber.join
} yield message

When a parent fiber joins a child fiber, it will succeed or fail in the same way as the child fiber, and the local states of the fibers will be merged.

Awaiting Fibers

Another method on Fiber is Fiber#await, which returns an effect containing an Exit value, which provides full information on how the fiber completed.

for {
fiber <- ZIO.succeed("Hi!").fork
exit <- fiber.await
} yield exit

Awaiting the exit values of fibers is different than joining them, because awaiting will not tie the fate of the parent fiber to that of the child fiber, and nor will it attempt to merge the local states of the fibers.

Interrupting Fibers

A fiber whose result is no longer needed may be interrupted, which immediately terminates the fiber, safely releasing all resources by running all finalizers.

Like await, Fiber#interrupt returns an Exit describing how the fiber completed.

for {
fiber <- ZIO.succeed("Hi!").forever.fork
exit <- fiber.interrupt
} yield exit

By design, the effect returned by Fiber#interrupt does not resume until the fiber has completed, which helps ensure your code does not spin up new fibers until the old one has terminated.

If this behavior (often called "back-pressuring") is not desired, you can ZIO#fork the interruption itself into a new fiber:

for {
fiber <- ZIO.succeed("Hi!").forever.fork
_ <- fiber.interrupt.fork // I don't care!
} yield ()

There is a shorthand for background interruption, which is the method Fiber#interruptFork.

Composing Fibers

ZIO lets you compose fibers with Fiber#zip or Fiber#zipWith.

These methods combine two fibers into a single fiber that produces the results of both. If either fiber fails, then the composed fiber will fail.

for {
fiber1 <- ZIO.succeed("Hi!").fork
fiber2 <- ZIO.succeed("Bye!").fork
fiber =
tuple <- fiber.join
} yield tuple

Another way fibers compose is with Fiber#orElse. If the first fiber succeeds, the composed fiber will succeed with its result; otherwise, the composed fiber will complete with the exit value of the second fiber (whether success or failure).

for {
fiber1 <-"Uh oh!").fork
fiber2 <- ZIO.succeed("Hurray!").fork
fiber = fiber1.orElse(fiber2)
message <- fiber.join
} yield message


ZIO provides parallel versions of many methods, which are named with a Par suffix that helps you identify opportunities to parallelize your code.

For example, the ordinary ZIO#zip method zips two effects together sequentially. But there is also a ZIO#zipPar method, which zips two effects together in parallel.

The following table summarizes some of the sequential operations and their corresponding parallel versions:

Zips two effects into oneZIO#zipZIO#zipPar
Zips two effects into oneZIO#zipWithZIO#zipWithPar
Zips multiple effects into oneZIO#tupledZIO#tupledPar
Collects from many effectsZIO.collectAllZIO.collectAllPar
Effectfully loop over valuesZIO.foreachZIO.foreachPar
Reduces many valuesZIO.reduceAllZIO.reduceAllPar
Merges many valuesZIO.mergeAllZIO.mergeAllPar

Because all these parallel operators return all the results, if any effect being parallelized fails, ZIO will automatically cancel the other running effects, as their results will not be used.

If the fail-fast behavior is not desired, potentially failing effects can be first converted into infallible effects using the ZIO#either or ZIO#option methods.


ZIO lets you race multiple effects concurrently, returning the first successful result:

for {
winner <- ZIO.succeed("Hello").race(ZIO.succeed("Goodbye"))
} yield winner

If you want the first success or failure, rather than the first success, then you can use left.either.race(right.either), for any two effects left and right.


ZIO has resource-safe, compositional timeouts that work on "small" effects, such as querying a database or calling a cloud API, or even "large" effects, such as running a streaming pipeline or fully handling a web request.

ZIO lets you timeout effects using the ZIO#timeout method, which returns a new effect that succeeds with an Option value.

A value of None indicates the timeout elapsed before the effect completed.


If an effect times out, then instead of continuing to execute in the background, it will be interrupted, for automatic efficiency.

Next Steps

If you are comfortable with basic concurrency, the next step is to learn about running effects.