Cost and Performance Guide

All observability views are non-materialized and query the Snowflake Event Table in real time. This means there are no additional storage costs within the Native App. However, every query you run consumes compute credits from your Snowflake warehouse, exactly as if you were querying the Event Table directly.

5.1 Cost Factors

Query compute costs scale with the following factors:

• Event volume – more active reasoners and workloads produce more events
• Time range queried – wider time ranges increase scan volume
• Query complexity – joins, aggregations, and multi-metric queries use more compute

5.2 Cost Optimization Strategies

Always apply time-range filters:

-- AVOID: Full table scan
SELECT * FROM
<rai_app_database>.observability_preview.logic_reasoner__memory_utilization;

-- RECOMMENDED: Bounded query
SELECT * FROM
<rai_app_database>.observability_preview.logic_reasoner__memory_utilization
WHERE timestamp >= DATEADD(hour, -24, CURRENT_TIMESTAMP());

Monitor your observability query costs:

SELECT query_id, query_text, total_elapsed_time, credits_used_cloud_services
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY
WHERE query_text ILIKE '%<rai_app_database>.observability%'
 AND start_time >= DATEADD(day, -7, CURRENT_TIMESTAMP())
ORDER BY total_elapsed_time DESC;