/plushcap/analysis/datastax/2019-08-performance-comparison-between-apache-pulsar-and-kafka-latency

Performance Comparison Between Apache Pulsar and Kafka: Latency

What's this blog post about?

Apache Pulsar provides consistent publishing latency over time, while Kafka shows periodic spikes in publishing latency. The 99th percentile publishing latency for Pulsar is essentially a straight line for the entire duration of the test, while Kafka has large spikes. Increasing the number of partitions lowers the publishing latency for Pulsar but increases it for Kafka. For end-to-end latency, Kafka beats Pulsar at the lowest partition count, but as the number of partitions is increased, Pulsar's average end-to-end latency approaches 3 ms while Kafka's increases to 11 ms. Disabling flushing to disk for Pulsar has little effect on its performance. In an apples-to-apples comparison with Kafka configured to flush each message to disk, Pulsar is clearly better in terms of latency. Reference(s): [Apache Pulsar](https://pulsar.apache.org/) [Apache Kafka](https://kafka.apache.org/)

Company
DataStax

Date published
Aug. 29, 2019

Author(s)
Chris Bartholomew

Word count
5062

Hacker News points
None found.

Language
English


By Matt Makai. 2021-2024.