Storage and Retention Costs
What you'll learn: How trace data volume scales with traffic and why balancing retention periods against storage costs is critical.
The Volume Challenge
Distributed tracing generates enormous amounts of data. Every request creates a trace with multiple spans—if you serve 1,000 requests per second across 10 services, that's potentially 10,000+ spans per second. At scale, this becomes millions of spans daily.
Unlike metrics (which aggregate) or logs (which you can filter aggressively), traces are inherently high-cardinality and span-heavy. Each unique request path, each tag combination, each span attribute adds storage burden.
The Cost Equation
Storage costs compound from two factors:
Volume = Traffic × Services × Spans per Service × Attributes per Span
Cost = Volume × Retention Period × Storage Price
If you retain 30 days of traces at full sampling, you're storing 30× more data than a 1-day retention window. Cloud storage pricing makes this expensive fast—especially when querying historical traces requires hot storage tiers.
Balancing Act
This is why sampling strategies (head-based, tail-based, adaptive, priority) exist—they reduce ingested volume. But you also need smart retention policies:
- Keep high-value traces (errors, slow requests) longer
- Age out normal traces aggressively
- Use cold storage for compliance-only retention
- Monitor your storage growth rate weekly
Think of traces like security camera footage: you can't store 4K video forever from every camera. You sample frame rates, compress, and archive strategically.
Key Takeaway: Trace storage costs scale with traffic and cardinality; effective cost management requires combining sampling strategies, tiered retention policies, and continuous monitoring of data growth.