apex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (APEXMALHAR-1963) Add abstract operator for Async Processing
Date Thu, 28 Jan 2016 12:09:39 GMT

    [ https://issues.apache.org/jira/browse/APEXMALHAR-1963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15121288#comment-15121288
] 

ASF GitHub Bot commented on APEXMALHAR-1963:
--------------------------------------------

Github user chinmaykolhatkar commented on a diff in the pull request:

    https://github.com/apache/incubator-apex-malhar/pull/157#discussion_r51114146
  
    --- Diff: library/src/main/java/com/datatorrent/lib/async/AbstractAsyncProcessor.java
---
    @@ -0,0 +1,341 @@
    +/**
    + * Licensed to the Apache Software Foundation (ASF) under one
    + * or more contributor license agreements.  See the NOTICE file
    + * distributed with this work for additional information
    + * regarding copyright ownership.  The ASF licenses this file
    + * to you under the Apache License, Version 2.0 (the
    + * "License"); you may not use this file except in compliance
    + * with the License.  You may obtain a copy of the License at
    + *
    + *   http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing,
    + * software distributed under the License is distributed on an
    + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
    + * KIND, either express or implied.  See the License for the
    + * specific language governing permissions and limitations
    + * under the License.
    + */
    +package com.datatorrent.lib.async;
    +
    +import java.util.Iterator;
    +import java.util.Queue;
    +import java.util.concurrent.ExecutorService;
    +import java.util.concurrent.Executors;
    +import java.util.concurrent.Future;
    +import javax.validation.constraints.Max;
    +import javax.validation.constraints.Min;
    +
    +import org.slf4j.Logger;
    +import org.slf4j.LoggerFactory;
    +
    +import com.google.common.collect.Lists;
    +
    +import com.datatorrent.api.Context;
    +import com.datatorrent.api.Operator.IdleTimeHandler;
    +import com.datatorrent.common.util.BaseOperator;
    +
    +
    +/**
    + * This base operator can be used to enqueue tuples for asynchronous and parallel processing
by separate worker threads.
    + *
    + * User can configure number of threads required for async processing. By default 32
threads will be created.
    + * These worker threads, spawned by operator thread, can also be configured to processing
timeout.
    + *
    + * <b>Workflow is as follows:</b>
    + * <ol>
    + *   <li>Implementing class can enqueue the tuples of type QUEUETUPLE using <i>enqueueTupleForProcessing</i>
    + *       method.</li>
    + *   <li>When a tuple gets its turn for execution, <i>processTupleAsync</i>
gets called
    + *       by worker thread. This is asynchronous call</li>
    + *   <li>Based on maintainTupleOrder parameter, the results are consolidated at
the end of every window and respective
    + *       call to <i>handleProcessedTuple</i> will be made. This callback
is in operator thread.</li>
    + * </ol>
    + *
    + *
    + * <b>Use case examples:</b>
    + * <ul>
    + *   <li>Parallel reads from external datastore/database system.</li>
    + *   <li>Operations which are long running tasks per tuple but DAG i/o should not
be blocked.</li>
    + * </ul>
    + *
    + * @param <QUEUETUPLE> input type of tuple
    + * @param <RESULTTUPLE> Result of async processing of tuple.
    + */
    +public abstract class AbstractAsyncProcessor<QUEUETUPLE, RESULTTUPLE> extends BaseOperator
implements IdleTimeHandler
    +{
    +  private static final Logger logger = LoggerFactory.getLogger(AbstractAsyncProcessor.class);
    +
    +  /**
    +   * This is an executor service which holds thread pool and enqueues task for processing.
    +   */
    +  private transient ExecutorService executor;
    +
    +  /**
    +   * Number of worker threads for processing
    +   * Minimum value is 1 and max allowed number of threads are 32.
    +   *
    +   * Default number of threads is 32.
    +   */
    +  @Min(1)
    +  @Max(32)
    +  private int numProcessors = 32;
    +
    +  /**
    +   * Tells whether the <i>handleProcessedTuple</i> should be called for tuples
in the same order in which
    +   * they get enqueued using <i>enqueueTupleForProcessing</i>.
    +   */
    +  private boolean maintainTupleOrder = false;
    +
    +  /**
    +   * Timeout in milliseconds after which processing of tuple may be terminated.
    +   */
    +  private long processTimeoutMs = 0;
    +
    +  /**
    +   * The queue holds following types of tuples:
    +   * <ul>
    +   *   <li>Yet to be processed</li>
    +   *   <li>In Process</li>
    +   *   <li>Completed processing but yet to be notified for process completion</li>
    +   * </ul>
    +   */
    +  private Queue<WorkItem> waitingTuples = Lists.newLinkedList();
    +
    +  /**
    +   * Setup method of this operator will initialize various types of executors possible.
    +   *
    +   * @param context OperatorContext
    +   */
    +  @Override public void setup(Context.OperatorContext context)
    +  {
    +    this.executor = Executors.newFixedThreadPool(numProcessors);
    +    replayTuplesIfRequired();
    +  }
    +
    +  /**
    +   * This method is called from setup of the operator. If <i>waitingTuples</i>
contains anything, the method enqueues
    +   * all those tuples for processing to worker threads.
    +   *
    +   * Dependening on condition the method will behave as follows:
    +   * <ul>
    +   *   <li>During fresh start of operator, this method will not have any effect
as there are no tuples in
    +   *       <i>waitingTuples</i>.</li>
    +   *   <li>During restore of operator to previous checkpointed state, if there
are any tuples already present
    +   *       in <i>waitingTuples</i> that means they're yet to be notified,
hence enqueued again.</li>
    +   * </ul>
    +   */
    +  private void replayTuplesIfRequired()
    +  {
    +    // Enqueue everything that is still left.
    +    // Its left because its either not completed processing OR its not emitted because
of ordering restrictions.
    +    for (WorkItem item : waitingTuples) {
    +      item.taskHandle = executor.submit(new Processor(item));
    +      item.processState = State.SUBMITTED;
    --- End diff --
    
    calling executor.submit after variables are set. This should take care of overrides.


> Add abstract operator for Async Processing
> ------------------------------------------
>
>                 Key: APEXMALHAR-1963
>                 URL: https://issues.apache.org/jira/browse/APEXMALHAR-1963
>             Project: Apache Apex Malhar
>          Issue Type: New Feature
>    Affects Versions: 3.3.0
>            Reporter: Chinmay Kolhatkar
>            Assignee: Chinmay Kolhatkar
>
> Create an abstract operator which does following:
> 1) Asynchronously processes the tuples
> 2) Have parallel executions
> 3) Make sure the outbound tuples are ordered similar to inbound tuples if configured.
> 4) Have processing timeout.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message