TechPrufer
Get started
Confluent
Confluent logoConfluent · Feature trail

Messaging System Performance Benchmarking

1 post · Aug 2020

What this trail covers

This feature thread tracks the evolution of understanding and comparing the performance characteristics of various messaging and event streaming systems. It covers detailed benchmarking methodologies, testbed configurations, and workload analysis to evaluate system throughput and latency. The goal is to provide engineers with data-driven insights to select the most appropriate system for their specific use cases, considering factors like hardware utilization, end-to-end latency, and tail latencies.

All Confluent trails
  1. Benchmarking RabbitMQ vs Kafka vs Pulsar PerformanceThis post introduces a comprehensive benchmarking framework (OpenMessaging Benchmark Framework - OMB) and methodology for comparing the performance of messaging systems like Kafka, RabbitMQ, and Pulsar. It details the testbed setup, including hardware, OS tuning, and memory configurations. The post presents detailed throughput and latency test results, highlighting Kafka's superior throughput and lower tail latencies compared to Pulsar and RabbitMQ. It also discusses the impact of durability mechanisms like fsync and replication on performance, and provides open-source access to the benchmarking code and collected metrics for reproducibility.