In summary, it is possible to create a new better TaskManger that takes
advantage of new concurrency libraries, but it cannot be directly
replaced by ThreadPoolExceutor, which is what Gregg is trying to say.
Fundamentally, we must learn everything about what we have and
understand it, in order to improve or replace it. I think we all are in
the process of understanding, but we must be careful not to be too
critical of each other, and be mindful not to hold our ground too
steadfast either, to remain humble, so the list doesn't descend into
argument.
I also think Chris has gotten to the heart of the matter:
The big feature that TaskManager has that Java 1.5 lacks is the ability
to reorder the tasks by asking each task if it wants to be before the
others (Task.runAfter()). That could be refactored into a custom
BlockingQueue implementation, I suppose.
Chris
It is interesting TPE doesn't create extra threads until the queue is
full.
Reading the source code from TaskManager it is apparent, maxThreads does
in fact limit the total number of threads in TaskManager and I was
previously mistaken.
This is taken directly from ThreadManager:
/** Add a new task. */
public synchronized void add(Task t) {
tasks.add(t);
boolean poke = true;
while (threads.size() < maxThreads && needThread()) {
Thread th;
try {
th = new TaskThread();
th.start();
<comment> A new TaskThread (which is a Thread) is created every time the
available threads is less than maxThreads AND a new thread is needed!
So even if a new Thread is needed, we cannot have more than
maxThreads</comment>
/** Return true if a new thread should be created (ignoring
maxThreads). */
protected boolean needThread() {
int bound = (int)(loadFactor * threads.size());
int max = tasks.size();
if (max < bound)
return false;
<comment> Above: This first piece of code shows that the load factor
influences the creation of new threads, in this case, the upper bound is
the load factor * the number of threads currently in use. So you must
have more tasks than threads * load factor, to get past this check.
This is similar to having the full Queue in TPE.</comment>
max--;
if (runAfter((Task)tasks.get(max), max))
return false;
<comment> Above: If the last task must be run after some other task
preceding it in the task list, return false.</comment>
int ready = firstPending + 1;
if (ready > bound)
return true;
<comment> Above: First pending is incremented every time a task is found
ready to run using "private boolean takeTask() " and it is decremented
every time a task is "public void run()" This may influence the
creation of a new thread, if there are more ready tasks than total
threads * loadFactor. </comment>
for (int i = firstPending; i < max; i++) {
if (!runAfter((Task)tasks.get(i), i)) {
ready++;
if (ready > bound)
return true;
}
<comment> Above this, piece of code loops over the tasks between
firstPending and the maximum bound of the tasks List. ready is
incremented every time a ready task is found. Then if there are more
tasks ready than the total number of threads * loadFactor, a new thread
will be created.
}
return false;
}
Gregg Wonderly wrote:
I am not against the thoughts here. TaskManager has a very important
roll that has specific needs. It is used to dispatch inbound calls
amongst other things. It is essential that it never fail to create a
thread to handle an inbound call. Without this behavior there is a
chance for distributed deadlock.
The key clause in TPE is that it says the extra threads are not
created until the queue is full. That is another form of "maximum".
If it has a documented behavior that has no maximum, then I am okay
with using it. I have just never seen it behave correctly based on
what I have experienced and so I am concerned about the maximum thread
issue.
There are of course many types of applications that can use a ceiling
on maximum threads to throttle things. But if you look at some of the
bug reports and discussions on the concurrency-interest list, there
are cases of this behavior popping up in the fork-join pool stuff too,
where the complexity of hoe work is divided and distributed creates
problems.
Gregg Wonderly
Sent from my iPhone
On Jun 5, 2010, at 8:25 AM, Dennis Reedy <[email protected]> wrote:
On Jun 5, 2010, at 609AM, Patrick Wright wrote:
One point I'd like to raise about using java.util.concurrent and TPE:
I think that over the long term, it makes sense to (re)use existing
utilities which are being maintained by domain experts rather than
custom utilities you've written yourself. The concurrent libraries
available since Java 5 were written and maintained by people widely
recognized to be very, very good at a very hard problem. That doesn't
mean they, or the library, is perfect, just that there is value in
building on their work and letting them take care of the bugs and
optimizations over time. The downside would be that if a River user
was stuck with, say, Java 5, they couldn't take advantage of bugfixes
or improvements in Java 6. On the other hand, that's true of the
entire JDK.
The max threads issue seems to me a non-issue. A JVM can allocate only
so many native threads before it runs out of OS resources; that's a
hard limit. You can set a max of Integer.MAX_VALUE but your VM would
die long, long before it reached that.
For me this is more of design policy decision. Re-use, intelligently
and selectively, where possible, to reduce your project's workload.
Patrick
+1
