Benchmarking Cloud Native Databases on Kubernetes - Performance and Tradeoffs

Learn how to benchmark cloud native databases in this informative conference talk. Explore the performance implications of running stateful applications and databases on Kubernetes, comparing various storage options including host, network, cloud, and cloud-native systems like Rook. Gain insights from PostgreSQL, CockroachDB, and filesystem benchmark results to make informed decisions about tradeoffs between performance, reliability, data retention, and manageability. Discover techniques for conducting your own benchmarks and understand the quantitative differences between storage configurations. Examine the benefits of bare metal platforms, analyze SSD throughput, and evaluate database performance across different setups. Leave with a comprehensive understanding of storage options and their impact on cloud native database performance in Kubernetes environments.

Introduction
Why run a database on Kubernetes
Why performance is critical
Types of storage
Latency and throughput
Micro benchmarks
Cobbler benchmarks
Benchmarking tips
Why use bare metal
Disclaimer
Bare Metal Platform
First set of tests
SSD throughput
Database benchmarks
Use local volumes
Shared network
Storage configuration
Results
Whats next
Conclusions
Questions