Serverless in Google Cloud: What’s Worth Exploring in 2025

[a_aleinikov]

Hi all,

I’d like to start a discussion on some of the most interesting serverless updates in Google Cloud this year. If anyone has hands-on experience, I’d appreciate hearing your thoughts.

Key updates:

• BigQuery with AI agents
  Natural language queries are now supported. I’ve tested it for reporting tasks — noticeably faster for analysts. Curious: how stable is it under heavy workloads?

• Serverless Spark in BigQuery
  A promising option for ETL and ML pipelines, especially for teams without deep Spark expertise. Has anyone run production workloads yet?

• Cloud Run + RAG + CI/CD
  Google is pushing Cloud Run for retrieval-augmented generation (RAG) use cases. Has anyone set up end-to-end CI/CD for these apps? Any pitfalls?

Discussion points:

1. Which serverless scenarios have proven most effective for you in 2025?

2. What bottlenecks (cold starts, observability, billing) still get in the way?

3. What monitoring or optimization tools do you rely on?

Looking forward to practical insights and lessons learned.

Best,
Aleksei

[mcbsalceda]

Hi @a_aleinikov,

BigQuery with AI agents: You might need to fine-tune things for heavier workloads, especially with more complex queries.

Serverless Spark in BigQuery: It’s still fairly new, so it might be good testing it under real production loads to check how it handles performance.

Cloud Run + RAG + CI/CD: Setting up a smooth CI/CD pipeline can be a bit tricky. You may want to use Cloud Build and Cloud Deploy to streamline your deployment process.

For monitoring and optimization, Cloud Monitoring and Cloud Logging are your go-to tools. Cold starts on Cloud Run can be managed by setting up "minimum instances." 

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.