Sudden Increase in Read Unit (PricingQuantity) Usage Since 25/06/2025 — Need Help Tracing the Cause

Support
, ,
1

Hi Pinecone Community,

I’ve noticed a sudden and consistent doubling in Read Unit (PricingQuantity) usage for one of our projects starting from June 25, 2025. There have been no changes on our end in terms of query volume, namespace size, vector dimensions, or metadata.

I’ve reviewed the usage data exported from the Pinecone dashboard and confirmed the increase is real and isolated to this project.

I’d appreciate any insights or suggestions on:

  • How to trace RU usage at a more granular level (project, namespace, index, or operation type)?
  • Whether there have been any recent backend changes or updates to the RU calculation model?
  • Best practices to diagnose and optimize usage in such cases?

Any guidance would be helpful — whether from Pinecone staff or community members who’ve faced similar issues.

Thanks in advance!

2

You can trace RU (Read Unit) usage at different levels using the following approaches:

  • Project-level usage: Open the Pinecone console, select your project, and review the usage metrics under the project’s usage section. This includes read units (RUs), write units (WUs), and storage, allowing you to track overall usage per project
    – Understandig cost
    – Rate limits .
  • Index-level and Namespace-level: Select a specific index to view detailed stats, including vector count and namespaces with their respective vector counts. When you query a namespace, the RU cost is determined by the namespace size, and each query response includes a usage field with the exact number of RUs consumed
    – GitHub, Read units .
  • Operation-level: Each query, fetch, or list operation returns the number of RUs used in the response. For example, after running a query, you can access response['usage']['read_units'] to see the RUs consumed for that operation.

Best practices to diagnose and optimize usage:

  • Monitor usage regularly: Use the console and API responses to track RUs at the project, index, and operation level
    See:
    – Managing cost guide
    – Read units demo
    – Rate limits .
  • Minimize namespace size for queries: Query only the relevant namespace(s) to reduce RU consumption, as the cost scales with namespace size, not the number of results
  • Optimize query parameters: Adjust parameters like top_k, include_metadata, and include_values. For example, increasing top_k does not increase RU cost unless you request metadata, which triggers additional fetch operations and thus more RUs .
  • Batch fetches: Fetching records in larger batches is more efficient, as 1 RU is used per 10 records fetched, with a minimum of 1 RU per request
  • Review rate limits: Be aware of RU rate limits per plan and per second to avoid throttling

For more information: