Java Concurrency – Understanding the Basics of Threads

https://turkogluc.com/java-concurrency-basics-of-threads/

by turkogluc • 6 years ago

142 82 6 years ago

Years ago, I tried learning how to use threads by following tutorials similar to this one, where you are taught how to implement threads from {python, java, c++}. However, it wasn't until I studied operating systems (when I returned to graduate school for computer science) was I able to wrap my mind around threads — from a language agnostic view point, how and what lightweight processes are, how to implement locks and synchronization barriers — and how they help facilitate concurrency.

selimthegrim• 6 years ago

What books are used in your program?

dbsmith83• 6 years ago

Idk about GP, but one book I highly recommend is Java Concurrency in Practice - https://www.amazon.com/Java-Concurrency-Practice-Brian-Goetz...

It's old, but the material holds up well since it covers a lot of fundamentals

derefr• 6 years ago

Seconded. It’s silly to learn threads “from the outside in” — thinking of them as an opaque abstraction and trying to understand the API they present. There’s no coherent abstraction there; you’ll only learn to cargo-cult the API, without gaining an intuition for what threads “are” or when and where you’d want to use those APIs.

The key thing to know, is that threads aren’t a first-class kernel object. In OS kernels, there are only OS processes and memory regions.

To learn about threads, you should just learn about OS processes; and then learn that distinct OS processes can share memory regions between them, often via subprocess-spawn-time inheritance. Learn what fork(2) does on POSIX, and how it manages to be fast.

Starting with that intuition, it’s simple to then absorb what “threads” actually are: a usage pattern for spawning and managing OS processes that share memory; and a set of convenience APIs (that may be in-kernel, as in Windows; or purely in userland, as in Linux) for setting up this usage pattern. Everything these “threading” APIs can do, you can do yourself directly using the process-management and memory-mapping APIs. And those same calls are all that e.g. libpthread is doing.

rusk• 6 years ago

I just visualise it as different instruction pointers with their own stack and shared heap. But I’m coming from Java so that might be an oversimplification!

formerly_proven5• 6 years ago

This strikes me as rather focused on Linux kernel implementation details, since in Windows processes and threads are actually distinct concepts (as opposed to the Linux kernel, which really only knows about tasks), where every live process has n>1 threads and the address space of threads is afaik strictly defined through the process it is part of.

djeiasbsbo• 6 years ago

I know about unix/linux processes, ipc and the relevant system calls (exec*, fork, clone, ...) but where do I continue from there?

Studying C, I haven't really come across threads other than trying out the things in `pthread.h`.

Would you recommend just reading the source code of that header for a better understanding?

yetkin2• 6 years ago

The problem is not the threads, it is the mutations of variables which boost the complexity of the code. So a tutorial on creation of theads actually an invitation to hell. Nothing is cool about it. Cool thing is achieving concurrency without threads/race conditions/shared memory

agumonkey• 6 years ago

Do you use other paradigm / languages ? (clojure comes to mind, but maybe others)

itronitron• 6 years ago

For Java at least, the Java Concurrency API is preferred.

mrjoelkemp• 6 years ago

Actor model with Elixir/Erlang and the BEAM VM.

abhishekjha4• 6 years ago

Akka actor framework comes to mind. I am in the process of learning it and it is definitely simpler to wrap your head around it.

rowls66• 6 years ago

I'll add Pony to the list. This language uses the actor model like Akka and Erlang, but allows for the safe sharing of data between actors enforced by an an ingenious use of the type system. The result is an actor programming model with better performance than Erlang because mutable data can be safely shared.

I have been a long time Java developer, and I have worked a lot with highly concurrent code. Pony really opened my eyes to what was possible.

Unfortunately, the language, standard library and runtime is still pretty immature. It does however have very good 'C' interop. So for some problems it would be a very good fit.

FpUser6• 6 years ago

The concepts of threads and concurrent data access is simple enough for any decent programmer to comprehend. There is no hell here. Sure there are some complex cases but complex cases will arise in many situations when programming things.

And achieving concurrency without shared memory is impossible in general case. Sure it is possible to isolate such access to a separate layer and make it transparent for the rest of the program but someone still has to program such layer.

JackFr• 6 years ago

The problem for novices is that a program that behaves correctly looks a lot like a correct program. Until one day it doesn’t.

And because you’re in production and getting random spurious failures, the panicked (but common) reaction is to wrap every shared resource in a synchronized block. Which makes an incorrect implementation worse but possibly correct.

MaxBarraclough• 6 years ago

The concepts of threads and concurrent data access is simple enough for any decent programmer to comprehend. There is no hell here.

It's notoriously difficult to reason about concurrent programs using intuition. Much more difficult than reasoning about non-concurrent imperative code. This is why there are articles like [0], and why a bug in a Wikipedia article on a fundamental concurrency algorithm went unnoticed until an analysis tool detected the issue, [1] and why lock-free algorithms in particular are so tricky to get right.

[0] https://www.cs.umd.edu/~pugh/java/memoryModel/DoubleCheckedL...

[1] [PDF] https://llvm.org/pubs/2008-08-SPIN-Pancam.pdf

mrfox321• 6 years ago

Can databases be efficiently implemented without shared access?

Can message passing accomplish this at the same level of performance?

although I agree that simplifying resource access should probably be considered before fully shared state.

mrkeen10• 6 years ago

Can databases be efficiently implemented without shared access?

Let me ask a different question: Why did databases take off in the way they did? Sure they persist stuff to disk, but so do files. What they offer is a concurrency model so good that you almost never think about it. Beginner programmers can competently write large, concurrent systems by writing single-threaded programs which are backed by a central DB, without even knowing the term "race condition".

If beginner database articles told users how to make database Threads, Thread groups, and how to signal and catch interruptions, I don't think databases would have enjoyed nearly as much popularity.

While Threads are fundamental to Java concurrency, I kinda agree with yetkin's point. It introduces the Thread footgun without even paying lip service to the problems of shared, mutable state.

Nursie2• 6 years ago

There's a lot cool about threads, and you can learn to implement them well.

Threads handled well do not need to have race conditions, and race conditions/deadlocks are also very possible in distributed, message-passing systems.

cle4• 6 years ago

A computer is a mutation machine, you cannot escape mutation by hand-waving it away. If you are writing programs in which you can achieve concurrency without threads and shared memory, it’s because you’re building on the shoulders of all the engineers who didn’t hand-wave it away. Many of us, due to product requirements, don’t have the luxury of using higher level abstractions like that.

mrkeen10• 6 years ago

And yet we're comfortable hand-waving GOTO away - that is, not calling computers GOTO machines.

Igelau• 6 years ago

All the cool people are in Hell. Only go to Heaven for the climate.

LgWoodenBadger• 6 years ago

With Futures and Executors/ExecutorServices I find that I rarely ever need to use raw Threads these days. Most of the thread-safety issues commonly encountered are eliminated with this approach as well.

alyandon• 6 years ago

Pretty much. I can't even recall the last time I've had to touch a low level threading class in Java since the executors cover so many of the common use cases.

Nursie2• 6 years ago

Likewise, with futures and executors I haven't had to touch threads directly for some time.

They give you the tools to just say "go away and do these things", which after years of dealing directly with pthreads in C was a breath of fresh air!

thaumasiotes• 6 years ago

But compare this very recent submission: https://news.ycombinator.com/item?id=24921657

which presents threads as the solution to the pain of using futures.

exabrial• 6 years ago

I was going to comment that as well. Futures, Managed Executors, tasks, runnables, are more higher level structures that are better suited for general use. Those constructs are often implemented using threads though, so it's worth knowing what's happening one layer below the abstraction layer.

silvestrov• 6 years ago

yeah, calling Thread.interrupt(), join() or similar methods are often a code smell for a bad "programming 101" teacher.

Executors are the way to go for almost all finite-time concurrency.

New threads should normally only be used for stuff that keeps running until the process quits.

cmckn• 6 years ago

Absolutely, executors offer much better (safer, more consistent) lifecycle management of a Runnable vs. a homegrown solution in my experience. The last time I extended Thread I think it was just to pull off a custom name format.

grandinj• 6 years ago

Except for that pesky swallowing of exceptions, I agree

turkogluc2• 6 years ago

I don't know why everyone is so tense in here :D I am planning to write a series of posts about concurrency (not because I am an expert but I want to improve myself by this way), including the executor framework, completable futures, locks and synchronisation. Obviously it is not possible to go into details of all in a single post. For one who knows operating system level details, and new to java I thing this is where you start, simply with Thread API. Even if you are using high-level abstraction this is these are basics to know. It is quite popular in the job interviews. And also it is quite common for people uses the Executor framework daily but does not know how to properly stop a thread, or non-daemon threads etc.

turkogluc2• 6 years ago

I have published a new article about Executor Framework: https://turkogluc.com/java-concurrency-executor-services/

danielhlockard• 6 years ago

The font colors in the code examples are quite hard to read on my machine. `public class Main {` is so dark that I have to highlight to read

cultus3• 6 years ago

Project Loom[0] is going to be coming out at some point. That brings direct JVM support for delimited continuations and fibers.

That's really going to change and simplify JVM concurrency and I think many other things. Delimited continuations can be used to implement algebraic effects, which is exciting for functional programming.

[0] https://openjdk.java.net/projects/loom/

cle4• 6 years ago

I honestly don’t think it will, not for a long time. It’s going to make things more complicated, because now we have to figure out how to move the entire ecosystem to it incrementally, while operating and maintaining systems during the long transition period, and making technical decisions at every step about how best to do concurrency with all those extra constraints.

cultus3• 6 years ago

I don't see it used much in legacy systems, but the benefits are large enough that I could see new Akka-ish libraries built on it.

an_opabinia• 6 years ago

Their compatibility story is alright though, at least they are trying to make "Thread" forward-compatible with the new runtime. And they are working on getting Netty to work, which will immediately get things compiling (not working) for a lot of projects.

The biggest challenge will be reconciling Futures, Netty and ThreadLocal (FiberLocal) patterns. Think defining a SQL transaction lifetime, a distributed lock, or a OpenTracing span. For Spring Framework people, no big deal. For everyone else, lots of complex decisions to make.

mrkeen10• 6 years ago

Delimited continuations can be used to implement algebraic effects, which is exciting for functional programming.

How-so? I thought the type system was the precluding factor for algebraic effects, not the threading model.

cultus3• 6 years ago

You won't be able to type them in Java, at least not now; however it would allow new JVM languages to do it. You could also do fibers on the JVM before, but they didn't have access to the actual JVM stack. That kind of limits their usefulness and performance.

JSavageOne• 6 years ago

I'm not a Java developer, but isn't RxJava the current best practice around managing concurrency in Java? I thought the consensus was that manually dealing with thread creation is too error-prone and unmanagable.

sk5t• 6 years ago

RxJava has some nice tools for async buffering, debounce, etc., but it pays to understand CountdownLatch, Semaphore, Mutex, ExecutorService, etc., and I would definitely not consider RxJava a substitute for other things.

Do avoid anything related to the hoary old Java "Future" class, though. CompletionStage or get out!

abhishekjha4• 6 years ago

We are using the Akka framework so as to not to have to deal with threads directly. Message passing and immutable objects simplify a lot while adding one more abstraction layer.

vips7L• 6 years ago

No, ExecutorServices are the current best practice around managing concurrency in Java. RxJava is only something you should use if you have specific performance requirements (with data to back it up), and you need the Observable pattern.

pjmlp• 6 years ago

RxJava got hyped in Android before Google went with Kotlt first, now they are into co-routines and depending on the Jetpack project, Java developers might still be able to use it, or be forced into Kotlin.

bassman9000• 6 years ago

Not quite. There were plenty and simpler abstractions to easily manage concurrency before, without even touching Thread. As someone has mentioned, Executors, ExecutorService, ThreadPools, ...

RxJava feels like an unnatural port from single-threaded mindsets.

abhishekjha4• 6 years ago

What a strange coincidence! I have started learning Java concurrency from the book[1]. I am on the synchronization chapter and it looks like managing threads via Runnable directly is going to be painful. I am hoping I get a good intro to Eecutors somewhere down the road. Adding the OP tutorial to go through once I have a better hold on writing concurrent programs.

1. https://www.amazon.in/Java-Threads-Concurrency-Utilities-Fri...

rurban• 6 years ago

And understand it's limitations: https://news.ycombinator.com/item?id=24955376 Java's insecure parallelism

edem• 6 years ago

I wouldn't accept anyone's PR with explicit Thread usage in them. You should either use some high-level construct like CompletableFuture or a concurrent data structure instead.

Java Concurrency – Understanding the Basics of Threads

Related Stories