JAVA スレッドプール | JAVA のスレッドプーリング

Javaスレッドプール ジョブを待機し、何度も再利用されるワーカースレッドのグループを表します。

スレッドプールの場合、固定サイズのスレッドのグループが作成されます。スレッドプールからスレッドが引き出され、サービスプロバイダーによってジョブが割り当てられます。ジョブの完了後、スレッドは再びスレッドプールに含まれます。

スレッドプールのメソッド

newFixedThreadPool(int s): このメソッドは、固定サイズのスレッドプールを作成します。

newCachedThreadPool(): このメソッドは、必要に応じて新しいスレッドを作成する新しいスレッドプールを作成しますが、使用できる場合は常に以前に作成されたスレッドを引き続き使用します。

newSingleThreadExecutor(): このメソッドは新しいスレッドを作成します。

Javaスレッドプールの利点

よりよい性能 新しいスレッドを作成する必要がないため、時間を節約できます。

リアルタイムの使用状況

これは、コンテナがリクエストを処理するためのスレッドプールを作成するサーブレットおよび JSP で使用されます。

Javaスレッドプールの例

ExecutorService と Executor を使用した Java スレッドプールの簡単な例を見てみましょう。

ファイル: WorkerThread.java

 import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; class WorkerThread implements Runnable { private String message; public WorkerThread(String s){ this.message=s; } public void run() { System.out.println(Thread.currentThread().getName()+&apos; (Start) message = &apos;+message); processmessage();//call processmessage method that sleeps the thread for 2 seconds System.out.println(Thread.currentThread().getName()+&apos; (End)&apos;);//prints thread name } private void processmessage() { try { Thread.sleep(2000); } catch (InterruptedException e) { e.printStackTrace(); } } }

ファイル: TestThreadPool.java

 public class TestThreadPool { public static void main(String[] args) { ExecutorService executor = Executors.newFixedThreadPool(5);//creating a pool of 5 threads for (int i = 0; i <10; i++) { runnable worker="new" workerthread('' + i); executor.execute(worker); calling execute method of executorservice } executor.shutdown(); while (!executor.isterminated()) system.out.println('finished all threads'); < pre> <p> <strong>Output:</strong> </p> <pre>pool-1-thread-1 (Start) message = 0 pool-1-thread-2 (Start) message = 1 pool-1-thread-3 (Start) message = 2 pool-1-thread-5 (Start) message = 4 pool-1-thread-4 (Start) message = 3 pool-1-thread-2 (End) pool-1-thread-2 (Start) message = 5 pool-1-thread-1 (End) pool-1-thread-1 (Start) message = 6 pool-1-thread-3 (End) pool-1-thread-3 (Start) message = 7 pool-1-thread-4 (End) pool-1-thread-4 (Start) message = 8 pool-1-thread-5 (End) pool-1-thread-5 (Start) message = 9 pool-1-thread-2 (End) pool-1-thread-1 (End) pool-1-thread-4 (End) pool-1-thread-3 (End) pool-1-thread-5 (End) Finished all threads </pre> download this example <h2>Thread Pool Example: 2</h2> <p>Let&apos;s see another example of the thread pool.</p> <p> <strong>FileName:</strong> ThreadPoolExample.java</p> <pre> // important import statements import java.util.Date; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; import java.text.SimpleDateFormat; class Tasks implements Runnable { private String taskName; // constructor of the class Tasks public Tasks(String str) { // initializing the field taskName taskName = str; } // Printing the task name and then sleeps for 1 sec // The complete process is getting repeated five times public void run() { try { for (int j = 0; j <= 5; j++) { if (j="=" 0) date dt="new" date(); simpledateformat sdf="new" simpledateformat('hh : mm ss'); prints the initialization time for every task system.out.println('initialization name: '+ taskname + '=" + sdf.format(dt)); } else { Date dt = new Date(); SimpleDateFormat sdf = new SimpleDateFormat(" hh execution system.out.println('time of is complete.'); } catch(interruptedexception ie) ie.printstacktrace(); public class threadpoolexample maximum number threads in thread pool static final int max_th="3;" main method void main(string argvs[]) creating five new tasks runnable rb1="new" tasks('task 1'); rb2="new" 2'); rb3="new" 3'); rb4="new" 4'); rb5="new" 5'); a with size fixed executorservice pl="Executors.newFixedThreadPool(MAX_TH);" passes objects to execute (step 3) pl.execute(rb1); pl.execute(rb2); pl.execute(rb3); pl.execute(rb4); pl.execute(rb5); shutdown pl.shutdown(); < pre> <p> <strong>Output:</strong> </p> <pre> Initialization time for the task name: task 1 = 06 : 13 : 02 Initialization time for the task name: task 2 = 06 : 13 : 02 Initialization time for the task name: task 3 = 06 : 13 : 02 Time of execution for the task name: task 1 = 06 : 13 : 04 Time of execution for the task name: task 2 = 06 : 13 : 04 Time of execution for the task name: task 3 = 06 : 13 : 04 Time of execution for the task name: task 1 = 06 : 13 : 05 Time of execution for the task name: task 2 = 06 : 13 : 05 Time of execution for the task name: task 3 = 06 : 13 : 05 Time of execution for the task name: task 1 = 06 : 13 : 06 Time of execution for the task name: task 2 = 06 : 13 : 06 Time of execution for the task name: task 3 = 06 : 13 : 06 Time of execution for the task name: task 1 = 06 : 13 : 07 Time of execution for the task name: task 2 = 06 : 13 : 07 Time of execution for the task name: task 3 = 06 : 13 : 07 Time of execution for the task name: task 1 = 06 : 13 : 08 Time of execution for the task name: task 2 = 06 : 13 : 08 Time of execution for the task name: task 3 = 06 : 13 : 08 task 2 is complete. Initialization time for the task name: task 4 = 06 : 13 : 09 task 1 is complete. Initialization time for the task name: task 5 = 06 : 13 : 09 task 3 is complete. Time of execution for the task name: task 4 = 06 : 13 : 10 Time of execution for the task name: task 5 = 06 : 13 : 10 Time of execution for the task name: task 4 = 06 : 13 : 11 Time of execution for the task name: task 5 = 06 : 13 : 11 Time of execution for the task name: task 4 = 06 : 13 : 12 Time of execution for the task name: task 5 = 06 : 13 : 12 Time of execution for the task name: task 4 = 06 : 13 : 13 Time of execution for the task name: task 5 = 06 : 13 : 13 Time of execution for the task name: task 4 = 06 : 13 : 14 Time of execution for the task name: task 5 = 06 : 13 : 14 task 4 is complete. task 5 is complete. </pre> <p> <strong>Explanation:</strong> It is evident by looking at the output of the program that tasks 4 and 5 are executed only when the thread has an idle thread. Until then, the extra tasks are put in the queue.</p> <p>The takeaway from the above example is when one wants to execute 50 tasks but is not willing to create 50 threads. In such a case, one can create a pool of 10 threads. Thus, 10 out of 50 tasks are assigned, and the rest are put in the queue. Whenever any thread out of 10 threads becomes idle, it picks up the 11<sup>th </sup>task. The other pending tasks are treated the same way.</p> <h2>Risks involved in Thread Pools</h2> <p>The following are the risk involved in the thread pools.</p> <p> <strong>Deadlock:</strong> It is a known fact that deadlock can come in any program that involves multithreading, and a thread pool introduces another scenario of deadlock. Consider a scenario where all the threads that are executing are waiting for the results from the threads that are blocked and waiting in the queue because of the non-availability of threads for the execution.</p> <p> <strong>Thread Leakage:</strong> Leakage of threads occurs when a thread is being removed from the pool to execute a task but is not returning to it after the completion of the task. For example, when a thread throws the exception and the pool class is not able to catch this exception, then the thread exits and reduces the thread pool size by 1. If the same thing repeats a number of times, then there are fair chances that the pool will become empty, and hence, there are no threads available in the pool for executing other requests.</p> <p> <strong>Resource Thrashing:</strong> A lot of time is wasted in context switching among threads when the size of the thread pool is very large. Whenever there are more threads than the optimal number may cause the starvation problem, and it leads to resource thrashing.</p> <h2>Points to Remember</h2> <p>Do not queue the tasks that are concurrently waiting for the results obtained from the other tasks. It may lead to a deadlock situation, as explained above.</p> <p>Care must be taken whenever threads are used for the operation that is long-lived. It may result in the waiting of thread forever and will finally lead to the leakage of the resource.</p> <p>In the end, the thread pool has to be ended explicitly. If it does not happen, then the program continues to execute, and it never ends. Invoke the shutdown() method on the thread pool to terminate the executor. Note that if someone tries to send another task to the executor after shutdown, it will throw a RejectedExecutionException.</p> <p>One needs to understand the tasks to effectively tune the thread pool. If the given tasks are contrasting, then one should look for pools for executing different varieties of tasks so that one can properly tune them.</p> <p>To reduce the probability of running JVM out of memory, one can control the maximum threads that can run in JVM. The thread pool cannot create new threads after it has reached the maximum limit.</p> <p>A thread pool can use the same used thread if the thread has finished its execution. Thus, the time and resources used for the creation of a new thread are saved.</p> <h2>Tuning the Thread Pool</h2> <p>The accurate size of a thread pool is decided by the number of available processors and the type of tasks the threads have to execute. If a system has the P processors that have only got the computation type processes, then the maximum size of the thread pool of P or P + 1 achieves the maximum efficiency. However, the tasks may have to wait for I/O, and in such a scenario, one has to take into consideration the ratio of the waiting time (W) and the service time (S) for the request; resulting in the maximum size of the pool P * (1 + W / S) for the maximum efficiency.</p> <h2>Conclusion</h2> <p>A thread pool is a very handy tool for organizing applications, especially on the server-side. Concept-wise, a thread pool is very easy to comprehend. However, one may have to look at a lot of issues when dealing with a thread pool. It is because the thread pool comes with some risks involved it (risks are discussed above).</p> <hr></=></pre></10;>

この例をダウンロードする

スレッドプールの例: 2

スレッドプールの別の例を見てみましょう。

ファイル名： ThreadPoolExample.java

 // important import statements import java.util.Date; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; import java.text.SimpleDateFormat; class Tasks implements Runnable { private String taskName; // constructor of the class Tasks public Tasks(String str) { // initializing the field taskName taskName = str; } // Printing the task name and then sleeps for 1 sec // The complete process is getting repeated five times public void run() { try { for (int j = 0; j <= 5; j++) { if (j="=" 0) date dt="new" date(); simpledateformat sdf="new" simpledateformat(\'hh : mm ss\'); prints the initialization time for every task system.out.println(\'initialization name: \'+ taskname + \'=" + sdf.format(dt)); } else { Date dt = new Date(); SimpleDateFormat sdf = new SimpleDateFormat(" hh execution system.out.println(\'time of is complete.\'); } catch(interruptedexception ie) ie.printstacktrace(); public class threadpoolexample maximum number threads in thread pool static final int max_th="3;" main method void main(string argvs[]) creating five new tasks runnable rb1="new" tasks(\'task 1\'); rb2="new" 2\'); rb3="new" 3\'); rb4="new" 4\'); rb5="new" 5\'); a with size fixed executorservice pl="Executors.newFixedThreadPool(MAX_TH);" passes objects to execute (step 3) pl.execute(rb1); pl.execute(rb2); pl.execute(rb3); pl.execute(rb4); pl.execute(rb5); shutdown pl.shutdown(); < pre> <p> <strong>Output:</strong> </p> <pre> Initialization time for the task name: task 1 = 06 : 13 : 02 Initialization time for the task name: task 2 = 06 : 13 : 02 Initialization time for the task name: task 3 = 06 : 13 : 02 Time of execution for the task name: task 1 = 06 : 13 : 04 Time of execution for the task name: task 2 = 06 : 13 : 04 Time of execution for the task name: task 3 = 06 : 13 : 04 Time of execution for the task name: task 1 = 06 : 13 : 05 Time of execution for the task name: task 2 = 06 : 13 : 05 Time of execution for the task name: task 3 = 06 : 13 : 05 Time of execution for the task name: task 1 = 06 : 13 : 06 Time of execution for the task name: task 2 = 06 : 13 : 06 Time of execution for the task name: task 3 = 06 : 13 : 06 Time of execution for the task name: task 1 = 06 : 13 : 07 Time of execution for the task name: task 2 = 06 : 13 : 07 Time of execution for the task name: task 3 = 06 : 13 : 07 Time of execution for the task name: task 1 = 06 : 13 : 08 Time of execution for the task name: task 2 = 06 : 13 : 08 Time of execution for the task name: task 3 = 06 : 13 : 08 task 2 is complete. Initialization time for the task name: task 4 = 06 : 13 : 09 task 1 is complete. Initialization time for the task name: task 5 = 06 : 13 : 09 task 3 is complete. Time of execution for the task name: task 4 = 06 : 13 : 10 Time of execution for the task name: task 5 = 06 : 13 : 10 Time of execution for the task name: task 4 = 06 : 13 : 11 Time of execution for the task name: task 5 = 06 : 13 : 11 Time of execution for the task name: task 4 = 06 : 13 : 12 Time of execution for the task name: task 5 = 06 : 13 : 12 Time of execution for the task name: task 4 = 06 : 13 : 13 Time of execution for the task name: task 5 = 06 : 13 : 13 Time of execution for the task name: task 4 = 06 : 13 : 14 Time of execution for the task name: task 5 = 06 : 13 : 14 task 4 is complete. task 5 is complete. </pre> <p> <strong>Explanation:</strong> It is evident by looking at the output of the program that tasks 4 and 5 are executed only when the thread has an idle thread. Until then, the extra tasks are put in the queue.</p> <p>The takeaway from the above example is when one wants to execute 50 tasks but is not willing to create 50 threads. In such a case, one can create a pool of 10 threads. Thus, 10 out of 50 tasks are assigned, and the rest are put in the queue. Whenever any thread out of 10 threads becomes idle, it picks up the 11<sup>th </sup>task. The other pending tasks are treated the same way.</p> <h2>Risks involved in Thread Pools</h2> <p>The following are the risk involved in the thread pools.</p> <p> <strong>Deadlock:</strong> It is a known fact that deadlock can come in any program that involves multithreading, and a thread pool introduces another scenario of deadlock. Consider a scenario where all the threads that are executing are waiting for the results from the threads that are blocked and waiting in the queue because of the non-availability of threads for the execution.</p> <p> <strong>Thread Leakage:</strong> Leakage of threads occurs when a thread is being removed from the pool to execute a task but is not returning to it after the completion of the task. For example, when a thread throws the exception and the pool class is not able to catch this exception, then the thread exits and reduces the thread pool size by 1. If the same thing repeats a number of times, then there are fair chances that the pool will become empty, and hence, there are no threads available in the pool for executing other requests.</p> <p> <strong>Resource Thrashing:</strong> A lot of time is wasted in context switching among threads when the size of the thread pool is very large. Whenever there are more threads than the optimal number may cause the starvation problem, and it leads to resource thrashing.</p> <h2>Points to Remember</h2> <p>Do not queue the tasks that are concurrently waiting for the results obtained from the other tasks. It may lead to a deadlock situation, as explained above.</p> <p>Care must be taken whenever threads are used for the operation that is long-lived. It may result in the waiting of thread forever and will finally lead to the leakage of the resource.</p> <p>In the end, the thread pool has to be ended explicitly. If it does not happen, then the program continues to execute, and it never ends. Invoke the shutdown() method on the thread pool to terminate the executor. Note that if someone tries to send another task to the executor after shutdown, it will throw a RejectedExecutionException.</p> <p>One needs to understand the tasks to effectively tune the thread pool. If the given tasks are contrasting, then one should look for pools for executing different varieties of tasks so that one can properly tune them.</p> <p>To reduce the probability of running JVM out of memory, one can control the maximum threads that can run in JVM. The thread pool cannot create new threads after it has reached the maximum limit.</p> <p>A thread pool can use the same used thread if the thread has finished its execution. Thus, the time and resources used for the creation of a new thread are saved.</p> <h2>Tuning the Thread Pool</h2> <p>The accurate size of a thread pool is decided by the number of available processors and the type of tasks the threads have to execute. If a system has the P processors that have only got the computation type processes, then the maximum size of the thread pool of P or P + 1 achieves the maximum efficiency. However, the tasks may have to wait for I/O, and in such a scenario, one has to take into consideration the ratio of the waiting time (W) and the service time (S) for the request; resulting in the maximum size of the pool P * (1 + W / S) for the maximum efficiency.</p> <h2>Conclusion</h2> <p>A thread pool is a very handy tool for organizing applications, especially on the server-side. Concept-wise, a thread pool is very easy to comprehend. However, one may have to look at a lot of issues when dealing with a thread pool. It is because the thread pool comes with some risks involved it (risks are discussed above).</p> <hr></=>

説明： プログラムの出力を見ると、スレッドにアイドル状態のスレッドがある場合にのみタスク 4 と 5 が実行されることがわかります。それまで、余分なタスクはキューに入れられます。

forループJava

上記の例からわかることは、50 個のタスクを実行したいが、50 個のスレッドを作成したくない場合です。このような場合、10 個のスレッドのプールを作成できます。したがって、50 個のタスクのうち 10 個が割り当てられ、残りはキューに入れられます。 10 個のスレッドのうちいずれかのスレッドがアイドル状態になると、11 個のスレッドが選択されます。^番目タスク。他の保留中のタスクも同様に処理されます。

スレッドプールに伴うリスク

スレッドプールに伴うリスクは次のとおりです。

デッドロック: マルチスレッドを伴うプログラムではデッドロックが発生する可能性があることは既知の事実であり、スレッドプールによってデッドロックの別のシナリオが発生します。実行用のスレッドが利用できないために、実行中のすべてのスレッドがブロックされ、キュー内で待機しているスレッドからの結果を待っているシナリオを考えてみましょう。

糸漏れ： スレッドのリークは、スレッドがタスクを実行するためにプールから削除されているが、タスクの完了後にスレッドがプールに戻らない場合に発生します。たとえば、スレッドが例外をスローし、プールクラスがこの例外をキャッチできない場合、スレッドは終了し、スレッドプールのサイズが 1 ずつ減ります。同じことが何度も繰り返される場合、かなりの確率で、プールは空になるため、他のリクエストを実行するためにプール内に使用できるスレッドはありません。

リソーススラッシング: スレッドプールのサイズが非常に大きい場合、スレッド間のコンテキストの切り替えに多くの時間が無駄になります。最適な数よりも多くのスレッドが存在すると、飢餓の問題が発生する可能性があり、リソースのスラッシングが発生します。

留意事項

他のタスクから得られる結果を同時に待機しているタスクをキューに入れないでください。上で説明したように、デッドロック状況が発生する可能性があります。

存続期間の長い操作にスレッドを使用する場合は常に注意が必要です。スレッドが永久に待機することになり、最終的にはリソースのリークにつながる可能性があります。

最終的には、スレッドプールを明示的に終了する必要があります。それが発生しない場合、プログラムは実行を継続し、終了することはありません。スレッドプールで shutdown() メソッドを呼び出して、エグゼキュータを終了します。シャットダウン後に誰かがエグゼキュータに別のタスクを送信しようとすると、RejectedExecutionException がスローされることに注意してください。

スレッドプールを効果的に調整するにはタスクを理解する必要があります。指定されたタスクが対照的である場合は、タスクを適切に調整できるように、さまざまな種類のタスクを実行するためのプールを探す必要があります。

JVM を実行するとメモリ不足になる可能性を減らすために、JVM で実行できる最大スレッドを制御できます。スレッドプールが最大制限に達すると、新しいスレッドを作成できなくなります。

スレッドが実行を終了している場合、スレッドプールは同じ使用済みスレッドを使用できます。したがって、新しいスレッドの作成に使用される時間とリソースが節約されます。

スレッドプールのチューニング

スレッドプールの正確なサイズは、使用可能なプロセッサの数と、スレッドが実行する必要があるタスクの種類によって決まります。システムに計算タイプのプロセスのみを取得する P 個のプロセッサがある場合、スレッドプールの最大サイズ P または P + 1 が最大の効率を実現します。ただし、タスクは I/O を待機する必要がある場合があり、そのようなシナリオでは、待機時間 (W) とリクエストのサービス時間 (S) の比率を考慮する必要があります。その結果、最大の効率を実現するためのプールの最大サイズ P * (1 + W / S) が得られます。

結論

スレッドプールは、特にサーバー側でアプリケーションを整理するための非常に便利なツールです。概念的には、スレッドプールは非常に理解しやすいです。ただし、スレッドプールを扱う場合は、多くの問題を検討する必要がある場合があります。これは、スレッドプールにはいくつかのリスクが伴うためです (リスクについては上で説明しています)。