GeoServer Performance Tuning Guide

00 Introduction

GeoServer is a powerful, open-source publishing platform for spatial data. Out of the box, it provides reasonable defaults, but production workloads require deliberate tuning. This guide covers the critical levers you can pull to optimize rendering speed, cache hit rates, database throughput, and resource utilization.

01 JVM & Memory Management

Java memory configuration is the single most impactful setting for GeoServer performance. Undersized heaps cause frequent garbage collection pauses, while oversized heaps increase GC cycle duration.

export CATALINA_OPTS="-Xms2g -Xmx4g -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:+UseStringDeduplication -XX:+AlwaysPreTouch"

• -Xms / -Xmx: Set initial and max heap to the same value to prevent runtime resizing pauses. Start with 2:1 ratio of physical RAM to heap size.
• G1GC: Garbage-First collector offers better predictable latency for medium-to-large heaps.
• String Deduplication: Reduces memory footprint of string-heavy WMS/WFS responses.

02 Database & Data Stores

Most GeoServer bottlenecks originate at the data store level. Ensure your spatial database is indexed, query-optimized, and connection-pooled.

• Create SP-GIST or BRIN indexes on geometry columns
• Run VACUUM ANALYZE regularly to update planner statistics
• Use HikariCP connection pooling (default in modern GeoServer) with maxLifetime=1800000
• Disable feature info queries (FeatureInfoFormats) when not needed—they are expensive

03 GeoWebCache & Tile Serving

Pre-rendering tiles via GeoWebCache (GWC) drastically reduces database load and improves map load times. Configure layer caching appropriately for your use case.

<!-- Enable disk-quota cleanup thread -->
<init-param>
  <param-name>disk-quota-manager-thread-interval</param-name>
  <param-value>30</param-value>
</init-param>

• Use image/webp or image/jpeg for raster tiles (better compression than PNG)
• Set appropriate Tile Size (256x256 is standard; 512x512 reduces HTTP requests)
• Enable BlobStore (S3/MinIO) for distributed tile storage and CDN integration

04 Threading & Web Container

GeoServer runs on Tomcat by default. Thread pool sizing determines concurrent request handling capacity. Too few threads cause queuing; too many cause context-switching overhead.

<Connector
  protocol="HTTP/1.1"
  port="8080"
  maxThreads="200"
  minSpareThreads="50"
  acceptCount="100"
/>

For high-concurrency environments, consider switching to Jetty or Undertow for better async I/O handling. Adjust org.geotools.shapefile.cache.size in web.xml if serving large vector layers.

05 Monitoring & Profiling

You cannot optimize what you cannot measure. Implement observability to track request latency, GC frequency, cache hit ratios, and database query times.

• Prometheus + Grafana: Export JMX metrics via jmx_prometheus_javaagent
• Logging: Set org.geotools and org.geoserver to INFO in production. Use DEBUG only for troubleshooting.
• Thread Dumps: Capture during load tests using jstack <pid> to identify blocking operations

06 Quick Optimization Checklist