Hi,
I am using storm kafka spout  (Version 1.1.0 ) and having multiple topics
configured using named subscription.
There are 5 topics in one group and each topic has 50 partitions.
What should be the spout parallelism to have maximum performance .

Is it 50*5 = 250 as the spout parallelism.Also auto commit for the kafka
spout is set as false .

Avro is the message format. Also any suggestions to improve the
deserialization performance.

*Code snippet for topology*

*config.put(Config.TOPOLOGY_EXECUTOR_RECEIVE_BUFFER_SIZE, 16384);*

*config.put(Config.TOPOLOGY_EXECUTOR_SEND_BUFFER_SIZE, 16384);*

*topicNames.forEach(name -> topicNameSet.add((String) name));*


* NamedSubscription namedSubscription = new
NamedSubscription(topicNameSet);*

*Builder<String, MdpData> builder     builder = new
KafkaSpoutConfig.Builder<String, MdpData>(kafkaBootstrapServers,
StringDeserializer.class, AvroDeserializer.class, namedSubscription);*

*  KafkaSpoutConfig<String, MdpData> spoutConfig =*

*
builder.setGroupId(kafkaGroupId).setProp(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG,
"false").setFirstPollOffsetStrategy(FirstPollOffsetStrategy.UNCOMMITTED_EARLIEST).build();*


Regards

Pradeep S

Reply via email to