Functions on the Run: Unified Serverless Platform
Exploring the unified serverless platform on Google Cloud with Cloud Functions and Cloud Run.
Events
| Name | Organizer | Date | Location | Attendees | Links |
|---|---|---|---|---|---|
| DevFest Cloud Manila 2024 | GDG Cloud Manila | 2024-10-05 | JY Campos Hall B, Bayanihan Center, Pasig, Philippines | 80 | 📊 Slide Deck |
QR Code for this session
Abstract
The serverless landscape on Google Cloud has reached a major milestone with the unification of its flagship products. Cloud Functions has officially been rebranded to Cloud Run functions, creating a single, powerful platform for both event-driven snippets and containerized microservices. This session dives into the implications of this change and explores how developers can now enjoy the best of both worlds: the simplicity of Functions-as-a-Service (FaaS) combined with the flexibility and advanced features of Cloud Run.
This session explores the technical benefits of this unified platform, including access to high-performance NVIDIA L4 GPUs for AI inference, custom volume mounts for Cloud Storage and Filestore, and advanced cost-optimization strategies like "Always-on CPU" and Committed Use Discounts (CUDs). The talk features a live demo building an AI-powered text summarization application using Gemma 2, Firebase Genkit, and Cloud Run, highlighting how to scale LLMs to zero when not in use. Whether participants are managing legacy Cloud Functions or building new AI-native applications, this session will provide a roadmap to master the next generation of serverless on Google Cloud.
Outline
- The evolution: Cloud Functions → Cloud Run functions
- What changed, what stayed the same
- New capabilities: GPU inference, volume mounts, Committed Use Discounts
- Cost optimization: Always-on CPU and scale to zero
- Live demo: AI text summarization with Gemma 2 + Firebase Genkit on Cloud Run
- Migration path for existing Cloud Functions workloads
Key Takeaways
- Cloud Functions is now Cloud Run functions - one unified serverless platform on Google Cloud
- FaaS simplicity and container flexibility are no longer a trade-off
- NVIDIA L4 GPU access enables on-demand AI inference without managing any infrastructure
- Scale-to-zero keeps LLM hosting costs near zero during idle periods
- CUDs make serverless economically competitive with always-on alternatives at scale