Scale smart, not just big: a practical guide to multi-node Amazon Timestream for InfluxDB 3 Enterprise

This article provides a practical guide to scaling Amazon Timestream for InfluxDB 3 Enterprise clusters, which now support up to 15 nodes with distinct roles for ingestion, querying, and compaction.

• Scale vertically (larger instances) for memory-bound queries, network bandwidth needs, or uniformly saturated clusters
• Scale horizontally (more nodes) for isolated bottlenecks, write throughput, fault tolerance, or role separation
• Reads and writes saturate independently; reads typically hit the ceiling first due to DataFusion pool and Parquet cache limits
• Adding dedicated query-only nodes isolates read contention from writes, improving performance without scaling all nodes
• Write throughput scales sub-linearly with ingest nodes due to S3 WAL flushes and compaction overhead
• Late-arriving data increases compaction work; consider dedicated compactor nodes if over 10% of data is out-of-order
• Decision table provided matching symptoms (slow queries, write backups, timeouts) to recommended scaling actions

Proper scaling decisions prevent cost waste and performance degradation by matching capacity additions to actual bottlenecks rather than scaling uniformly.