1

What I'm trying to accomplish is I have a action block with MaxDegreeOfParallelism = 4. I want to create one local instance of a session object I have for each parallel path, So I want to total of 4 session objects. If this was threads I would creating something like:

 ThreadLocal<Session> sessionPerThread = new ThreadLocal<Session>(() => new Session());

I know blocks are not threads so I'm looking for something similar but for blocks. Any way to create this?

This block is in a service and runs for months on end. During that time period tons of threads are used for each concurrent slot of the block so thread local storage is not appropriate. I need something tied to the logical block slot. Also this block never completes, it runs the entire lifetime of the service.

Note: The above suggested answer is not valid for what I am asking. I'm specifically asking for something different than thread local and the above answer is using thread local. This is a different question entirely.

Aaron Stainback
  • 3,489
  • 2
  • 31
  • 32
  • possible duplicate of [TPL Data Flow Thread Local Data](http://stackoverflow.com/questions/14693452/tpl-data-flow-thread-local-data) – svick Mar 07 '13 at 16:47
  • No this is not a duplicate of TPL Data Flow Thread Local Data. They are asking on how to use Thread Local Data, this is explicitly what I don't want to use. I want something that is tied to concurrent block slots instead not threads. Any given slot in a block can cross many threads over the lifetime of a block, this is not what I want. – Aaron Stainback Mar 07 '13 at 22:51
  • Except there dataflow blocks have no concept of “slots”. There is max degree of parallelism, but that's not the same, there isn't a slot that would travel between threads. Could you explain why exactly do you want to do it this way? – svick Mar 07 '13 at 23:18
  • I have a session object that is really big in size and takes a very long time to create, several seconds. I want to have four of these used concurrently but any one of them must not be used concurrently it's not thread safe. I will keep them around for the lifetime of the service. – Aaron Stainback Mar 08 '13 at 01:00

1 Answers1

4

As it sounds like you already know, Dataflow blocks provide absolutely no guarantee of correlation between blocks, execution, and threads. Even with max parallelism set to 4, all 4 tasks could be executing on the same thread. Or an individual task may execute on many threads.

Given that you ultimately want to reuse n instances of an expensive service for your n degrees of parallelism, let's take dataflow completely out of the picture for a minute, since it doesn't help (or directly hinder) you from any general solution to this problem. It's actually quite simple. You can use a ConcurrentStack<T>, where T is the type of your service that is expensive to instantiate. You have code that appears at the top of the method (or delegate) that represents one of your parallel units of work:

private ConcurrentStack<T> reusableServices;

private void DoWork() {
    T service;
    if (!this.reusableServices.TryPop(out service)) {
        service = new T(); // expensive construction
    }

    // Use your shared service.
    //// Code here.

    // Put the service back when we're done with it so someone else can use it.
    this.reusableServices.Push(service);
}

Now in this way, you can quickly see that you create exactly as many instances of your expensive service as you have parallel executions of DoWork(). You don't even have to hard-code the degree of parallelism you expect. And it's orthogonal to how you actually schedule that parallelism (so threadpool, Dataflow, PLINQ, etc. doesn't matter).

So you can just use DoWork() as your Dataflow block's delegate and you're set to go.

Of course, there's nothing magical about ConcurrentStack<T> here, except that the locks around push and pop are built into the type so you don't have to do it yourself.

Andrew Arnott
  • 80,040
  • 26
  • 132
  • 171
  • Yes I am familiar with this pattern but I was looking for some way to do this directly with blocks. I have created object pools in a similar fashion but using ConcurrentBag. I'm going to give it some more time to see if anyone knows of a way to do it directly with blocks if not, I'll mark yours as the answer. – Aaron Stainback Mar 08 '13 at 04:59
  • How could 4 synchronous `Task`s execute on the same thread (assuming they actually execute in parallel, not just in series). Or did you means asynchronous `Task`s? – svick Mar 08 '13 at 08:22
  • @svick Dataflow cannot guarantee parallel execution. All it can do is schedule *n* tasks to the designated TaskScheduler at once. At that point, it's up to the TaskScheduler to determine when the tasks actually execute. Suppose the TaskScheduler was actually one that executes everything on the STA thread of a windowing app. Well, in that case all "parallel" tasks get scheduled at once, but only executed one at a time, on the same STA thread. If the TaskScheduler were the default threadpool one, the threadpool may still execute all of them on the same thread - not usually, but it can happen. – Andrew Arnott Mar 09 '13 at 14:49
  • Aaron, there isn't. But sure, we can wait. :) – Andrew Arnott Mar 09 '13 at 14:49
  • @AndrewArnott Sure, but I don't see why that would be a problem. I think the requirement here isn't to use exactly 4 local objects, it's to use *at most* 4 objects. – svick Mar 09 '13 at 15:02
  • The question said "I want to total of 4 session objects". But it's not the # of objects that interests me. In fact my solution above doesn't guarantee 4, but rather as many as necessary given the degree of parallelism actually achieved. – Andrew Arnott Mar 09 '13 at 18:49
  • BTW @svick, I missed that you said "assuming they actually execute in parallel" in your earlier comment. Even then, there are ways to execute synchronous work in parallel even on the same thread (well, sort of interleaved rather than strictly parallel, but it works out just about the same). See [Fibers](http://en.wikipedia.org/wiki/Fiber_(computer_science)) which actually are available in .NET if you have the library. – Andrew Arnott Mar 09 '13 at 18:51