Project Strawberry: GDS On Demand Compute
summary
Project Strawberry runs Neo4j Graph Data Science as short‑lived GDS compute sessions hosted in Aura so GDS runs off the database JVM on separate machines, avoiding heap and CPU contention. Current Python client demo projects 3.1M nodes and 17M relationships via remote projection in ~20s, starts a session in ~28s, runs PageRank in ~3.7s, writes results back in ~10s, and supports TTL with a hard 30‑day expiry. Early access is open to Aura virtual dedicated cloud customers.
AI-written summary · about
chapters (16)
- 0:00Introduction to Project Strawberry
- 1:58Graph Data Science recap
- 3:45Challenges with current GDS setup
- 6:18Project Strawberry goals
- 7:44Demo introduction
- 8:06Hands-on session setup
- 10:15Creating a GDS session
- 12:24Projecting graph data remotely
- 13:49Running PageRank algorithm
- 15:16Running FastRP embedding
- 16:43Inspecting and writing back results
- 18:08Deleting the session
- 18:29Self-managed database solution
- 19:12Current features overview
- 20:37Early access program details
- 21:41Future ideas and Q&A
related talks
24mGDS Agent for Graph Algorithmic Reasoning
30mNeurosymbolic Architecture in Production with Neo4j, GDS, Node2Vec & LLM Grounding
16mPreview: Cloud Computing
47mCloud Native and Sustainable, Reproducible Scientific Computing
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...
18mArchitecting Multi-tenant Data-center Networks for GPU Customers
01h 17mGraph Database and Retrieval Augmented Generation
06mPlanet scale 2.0: Cloud games done right