# Performance Tuning

# Performance Tuning

Optimizing your Curiosity Workspace ensures fast search responses and efficient data ingestion, especially as your graph scales.

# Ingestion Performance

• Batch Sizes: Use optimal batch sizes (typically 100-500 items) when committing to the graph.
• Parallelism: Run multiple ingestion tasks in parallel if the source system and network bandwidth allow.
• Skip Unchanged: Implement logic to skip records that haven't changed since the last sync.

# Search Optimization

• Indexing: Only index fields that are actually used for searching or filtering.
• Query Complexity: Simplify complex graph traversals where possible.
• Caching: Leverage endpoint caching for frequently requested data.

# Resource Allocation

• Memory: Ensure sufficient RAM for the graph engine and search indexes.
• CPU: Monitor CPU usage during heavy ingestion or complex AI workflows.
• Disk I/O: Use high-performance SSDs for storage to minimize latency.

# Scaling Strategies

As your data grows, consider:

• Vertical Scaling: Increasing CPU and RAM for the workspace service.
• Horizontal Scaling: Distributing search and ingestion load across multiple instances (for enterprise deployments). Curiosity supports a primary + read-replicas model for horizontal scalability.

# Monitoring for Bottlenecks

Regularly check the Monitoring dashboard for:

• High query latency
• Slow ingestion rates
• Memory pressure
• Failed background tasks