Table of Contents

Best practices for deployment configuration

Follow these guidelines for the recommended deployment configuration.

Expected throughput

Data throughput depends on the number of nodes, CPUs, and partitions, as well as the replication factor and bandwidth.

Review the results of tests on three running stream service nodes on machines with the following configuration:

  • CPU cores: 2
  • Memory (GB): 8
  • Bandwidth (Mbps): 450
  • Number of partitions: 20
  • Replication factor: 2

The following table shows the test results for writing messages to the stream (producer):

Records Record
Threads Throughput
5000000 100 1 83675 2.2 8.4
    5 172276.1 11.9 17.2
    10 216682.8 40.4 21.7
  100 1 32967.7 5.1 33
    5 53033.7 48.6 53
    10 49861.7 174.3 49.8
  100 1 76812.1 2.4 7.7
    5 165317.4 13.3 16.5
    10 203216.3 46.1 19.38
  1000 1 35865.7 4.8 35.9
    5 52456.7 41.9 52.4
    10 50266.1 158.8 50.3

Producer throughput - test results

The following table presents the test results for reading messages from the stream (consumer):

For the consumer, the replication factor is not important because the consumer reads from the leading partition.
Records Record size
Threads Throughput
5000000 100 1 120673.8 12
    5 150465 15
    10 143395.3 14.3
  1000 1 54128.4 54.1
    5 55903.3 55.9
    10 54674.5 54.7

Consumer throughput - test results

Disk space requirements

By default, the Kafka cluster stores data for seven days. You can change that time by overriding the log.retention.hours property.

For more information, see the “Overriding Kafka settings” section in the Kafka common configuration options article.


Your goal is to process 100,000 messages per second, 500 bytes each, and to keep messages on the disk for one day. The replication factor is set to 2.

The expected throughput is 50MB/sec:

  • 3GB is used in one minute for a single copy of the data.
  • 6GB of disk space is used in one minute due to the replication factor of 2.
  • The total throughput is 360GB in one hour and 8.64TB in one day.
  • Apart from your data, the Kafka cluster uses additional disk space for internal data (around 10% of the data size).

In that sample scenario, the total minimal disk size should be 9.5TB.


Depending on your needs, you can choose data compression using one of the algorithms that Kafka supports: GZIP, Snappy, or LZ4. Consider the following aspects:

  • GZIP requires less bandwidth and disk space, but this algorithm might not saturate your network while the maximum throughput is reached.
  • Snappy is much faster than GZIP, but the compression ratio is low, which means that throughput might be limited when the maximum network capacity is reached.
  • LZ4 maximizes the performance.

Review the following table and diagram with throughput and bandwidth usage per codec:

Codec Throughput % Bandwidth %
none 100 100
gzip 47.5 5.2
snappy 116.3 64.1
lz4 188.9 34.5

Throughput and bandwidth per codec (%)


Throughput and bandwidth metrics per codec (%)


Disk usage metrics per codec

To define the compression algorithm, update the compression.type property. By default, no compression is applied.

    Data folder

    By default, the stream service keeps its data in the current working directory. For Apache Tomcat, the working directory is located in: <your_tomcat_folder>/kafka-data

    Ensure that you set up that path to the location with enough disk space. For more information, see the “Kafka commit log location” section of the Kafka common configuration options article.


    To view the main outline for this article, see Kafka as a streaming service.

    Suggest Edit

    33% found this useful

    Have a question? Get answers now.

    Visit the Collaboration Center to ask questions, engage in discussions, share ideas, and help others.