Improving Cognite Functions stability with new rate limits

Hi there,

We're introducing new rate limits for Cognite Function Calls, affecting less than 5% of CDF projects based on current usage patterns.

Cognite Functions has seen significant growth, and we've identified usage patterns with very high running call counts that are impacting system capacity and performance for all users. Our new rate limits will ensure more equitable resource distribution and maintain reliable service quality across all CDF projects.

What's changing:

We're introducing limits on concurrent function calls with a phased rollout:

• Current: Undocumented concurrent running call limits (based on underlying cloud provider)

• New: Maximum of 250 concurrent running calls per CDF project (independent of cloud provider)

• New projects: Limits apply immediately starting with this announcement

• Existing projects: Limits take effect November 1, 2025

• Additionally, we're updating the HTTP response status code for the case where the limit of 100 concurrent running function calls per function is exceeded. The code will change from 400 to 429, aligning with HTTP standards.

What this means for you:

• Most CDF projects won't notice any difference. The 250 concurrent running call limit supports typical usage patterns across most workloads.

• CDF projects currently exceeding this limit will receive direct outreach from our team with optimization guidance and transition support.

• We recommend reviewing your function concurrency patterns and consider optimizations like batching or queue management for high-volume scenarios.

• Upon reaching the limit, new function call requests will return HTTP 429 (Too Many Requests), signaling clients to back off and retry later. Some SDKs like the official Python SDK already handle automatic retries with exponential backoff for 429 responses, minimizing the need for code changes.

We take these decisions seriously. We're committed to supporting high-concurrency use cases through architectural improvements and will continue to monitor usage patterns to ensure these limits serve our community effectively.

We also recognize the importance of maintaining performance during this transition. Our team will be closely monitoring system behavior and is ready to provide direct support for projects that need assistance with optimization strategies.