You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by 林海涛(信息技术部交易云技术研发组) <li...@gf.com.cn> on 2016/06/01 03:05:55 UTC

Re: Re: How to improve the intercommunication latency of spout/bolt

Thank you. Is there any other optimizing method such as modifying storm config? I set the TOPOLOGY_DISRUPTOR_BATCH_SIZE and TOPOLOGY_DISRUPTOR_BATCH_TIMEOUT_MILLIS to 1, it seems better only when I set the worker number to 1.

发件人: Kevin Conaway <ke...@gmail.com>>
答复: "user@storm.apache.org<ma...@storm.apache.org>" <us...@storm.apache.org>>
日期: 2016年5月31日 星期二 下午8:53
至: "user@storm.apache.org<ma...@storm.apache.org>" <us...@storm.apache.org>>
主题: Re: How to improve the intercommunication latency of spout/bolt

Try using localOrShuffle grouping.  Storm will attempt to pass messages directly to the next component within the same JVM, if possible

On Tuesday, May 31, 2016, 林海涛 <li...@gf.com.cn>> wrote:
Hello.
I do test with a simple topology to test the intercommunication latency of spout/bolt. It’s just emit the current nano timestamp from a spout and print the time difference when a bolt receive it.
I deploy my storm cluster in my own machine with docker container (one nimbus, one supervisor), and run the topology in cluster mode.
code as below:


public classRandomSpout extends BaseRichSpout{

SpoutOutputCollector _collector;


  publicvoid open(Mapconf, TopologyContext context, SpoutOutputCollector collector) {

    _collector = collector;

  }



  publicvoid nextTuple() {

    Utils.sleep(1000);

longcurrentTime = System.nanoTime();

    _collector.emit(new Values(currentTime));



  }


publicvoid declareOutputFields(OutputFieldsDeclarer arg0) {

// TODO Auto-generated method stub

arg0.declare(new Fields("value"));

}

}


public classPrintBolt extends BaseRichBolt{

private LogFileWriter _logFile;


publicvoid execute(Tuple arg0) {

// TODO Auto-generated method stub

longprevTime = arg0.getLong(0);

longcurrentTime = System.nanoTime();

_logFile.writeLog("cost: " + (currentTime - prevTime));

}


publicvoid prepare(Maparg0, TopologyContext arg1, OutputCollector arg2) {

// TODO Auto-generated method stub

try {

_logFile = new LogFileWriter("StormTest”, this.getClass().getSimpleName());

} catch (Exception e) {

// TODO Auto-generated catch block

e.printStackTrace();

}

}


publicvoid declareOutputFields(OutputFieldsDeclarer arg0) {

// TODO Auto-generated method stub

arg0.declare(new Fields("value"));

}


}


public class Topology

{

    public staticvoid main( String[] args )

    {

    TopologyBuilder builder = new TopologyBuilder();

    builder.setSpout("spout", new RandomSpout(), 1);

    builder.setBolt("bolt", new PrintBolt(), 1).shuffleGrouping("spout");



    Config conf = new Config();

    conf.setDebug(false);

    if(args.length > 0){

            // cluster submit.

    conf.setNumWorkers(2);

        conf.setNumAckers(0);

            try {

                StormSubmitter.submitTopology("stormTest", conf, builder.createTopology());

            } catch (Exception e) {

e.printStackTrace();

    }

        }else{

            new LocalCluster().submitTopology("stormTest", conf, builder.createTopology());

        }

    }

}


Output is below:

[2016-05-31 09:13:53]cost: 1960336
[2016-05-31 09:13:54]cost: 2600239
[2016-05-31 09:13:55]cost: 3103449
[2016-05-31 09:13:56]cost: 3206544
[2016-05-31 09:13:57]cost: 3783647
[2016-05-31 09:13:58]cost: 3635923
[2016-05-31 09:13:59]cost: 3887787
[2016-05-31 09:14:00]cost: 1623692
[2016-05-31 09:14:01]cost: 2524674
[2016-05-31 09:14:02]cost: 3383506
[2016-05-31 09:14:03]cost: 3898478
[2016-05-31 09:14:04]cost: 2120949
[2016-05-31 09:14:05]cost: 3756272
[2016-05-31 09:14:06]cost: 2877997
[2016-05-31 09:14:07]cost: 3432532
[2016-05-31 09:14:08]cost: 3638306
[2016-05-31 09:14:09]cost: 2958907
[2016-05-31 09:14:10]cost: 2742666
[2016-05-31 09:14:11]cost: 3024576
[2016-05-31 09:14:12]cost: 2822562
[2016-05-31 09:14:13]cost: 2623060
[2016-05-31 09:14:14]cost: 4045938

Obviously, there is a 2ms latency approximately. It seems not good for me. How can I reduce the latency?


--
Kevin Conaway
http://www.linkedin.com/pub/kevin-conaway/7/107/580/
https://github.com/kevinconaway