[
https://issues.apache.org/jira/browse/KAFKA-4474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15748575#comment-15748575
]
Juan Chorro commented on KAFKA-4474:
------------------------------------
Hi again!
We have been doing more performance tests and we have observed an anomaly
behavior. In link below you can see all gotten information:
https://docs.google.com/spreadsheets/d/1iVKIp3vGCZKSlByaaEhM5hiA3cNA75zzNGdMNeIMO68/edit?usp=sharing
In all cases each one service has its own node. We have gotten the consumer and
producer throughtput by JMX protocol with jconsole tool.
We have a synthetic producer that injects ~100K messages per second to
kafka-streams application and we can seei next cases:
* Case A: We have 1 zookeeper server, 1 kafka-broker and 1 kafka-streams app.
Also we have two topics, input and output, with 4 partitions each one. We have
a synthetic producer that injects ~100K messages per second in input topic, the
kafka-streams app is consuming ~20K messages per second and in output topic the
app is producing to ~4K messages per second. Where is others ~16K messages per
second whether I don't observe a excesive RAM increment?
* Case B: We have 1 zookeeper server, 2 kafka-brokers and 1 kafka-streams app.
We have two topics, input and output, with 2 partitions each one. We have a
synthetic producer that injects ~100K messages per second in input topic, the
kafka-streams app is consuming ~100K messages per second and in output topic
the app is producing to ~100K messages per second. This case is correct!
* Case C: We have 1 zookeeper server, 2 kafka-brokers and 1 kafka-streams app.
We have two topics, input and output, with 4 partitions each one. We have a
synthetic producer that injects ~100K nessages per second in input topic, the
kafka-streams app is consuming ~20K messages per second and in output topic the
app is producing to ~4K messages per second. This case is the same that Case A
but with different #kafka-brokers.
* Case D: We have 1 zookeeper server, 4 kafka-brokers and 1 kafka-streams app.
We have two topics, input and output, with 4 partitions each one. We have a
synthetic producer that injects ~100K messages per second in input topic, the
kafka-streams app is consuming ~820K messages per second and in output topic
the app is producing to ~100K messages per second. In this case as synthetic
producer as kafka-streams producer have the same throughput but the
kafka-streams consumer gets ~820K messages per second and I don't know why.
I don't understand why from Case B to Case C the consumer's throughput is lower
when increase #partitions.
Do you like same like me?
Do I have some wrong concepts?
If you need anything else, feel free asking me for it.
> Poor kafka-streams throughput
> -----------------------------
>
> Key: KAFKA-4474
> URL: https://issues.apache.org/jira/browse/KAFKA-4474
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Affects Versions: 0.10.1.0
> Reporter: Juan Chorro
> Assignee: Eno Thereska
> Attachments: hctop sreenshot.png
>
>
> Hi!
> I'm writing because I have a worry about kafka-streams throughput.
> I have only a kafka-streams application instance that consumes from 'input'
> topic, prints on the screen and produces in 'output' topic. All topics have 4
> partitions. As can be observed the topology is very simple.
> I produce 120K messages/second to 'input' topic, when I measure the 'output'
> topic I detect that I'm receiving ~4K messages/second. I had next
> configuration (Remaining parameters by default):
> application.id: myApp
> bootstrap.servers: localhost:9092
> zookeeper.connect: localhost:2181
> num.stream.threads: 1
> I was doing proofs and tests without success, but when I created a new
> 'input' topic with 1 partition (Maintain 'output' topic with 4 partitions) I
> got in 'output' topic 120K messages/seconds.
> I have been doing some performance tests and proof with next cases (All
> topics have 4 partitions in all cases):
> Case A - 1 Instance:
> - With num.stream.threads set to 1 I had ~3785 messages/second
> - With num.stream.threads set to 2 I had ~3938 messages/second
> - With num.stream.threads set to 4 I had ~120K messages/second
> Case B - 2 Instances:
> - With num.stream.threads set to 1 I had ~3930 messages/second for each
> instance (And throughput ~8K messages/second)
> - With num.stream.threads set to 2 I had ~3945 messages/second for each
> instance (And more or less same throughput that with num.stream.threads set
> to 1)
> Case C - 4 Instances
> - With num.stream.threads set to 1 I had 3946 messages/seconds for each
> instance (And throughput ~17K messages/second):
> As can be observed when num.stream.threads is set to #partitions I have best
> results. Then I have next questions:
> - Why whether I have a topic with #partitions > 1 and with
> num.streams.threads is set to 1 I have ~4K messages/second always?
> - In case C. 4 instances with num.stream.threads set to 1 should be better
> that 1 instance with num.stream.threads set to 4. Is corrects this
> supposition?
> This is the kafka-streams application that I use:
> https://gist.github.com/Chorro/5522ec4acd1a005eb8c9663da86f5a18
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
