We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hacker News·14d ago·5 min read
Slater – Low-memory graphdb designed for read-heavy graphs
TL;DR
Slater is a low-memory graph database that efficiently handles large graphs, allowing for hundreds of millions of nodes and billions of edges to be served with minimal RAM usage. It supports standard Neo4j drivers and integrates disk-native vector search alongside live, durable writes.
✦ Why It Matters
Consider using Slater for your next project involving large-scale graph data to optimize memory usage.
Key Takeaways
- Utilize Slater to serve large graphs without high memory costs.
- Leverage live writes without impacting read performance.
- Integrate vector search capabilities alongside graph queries.
- Deploy easily with built-in security and multi-tenancy features.
How It Works
Slater compiles graphs into a content-addressed on-disk image, allowing it to serve large datasets by paging data from disk as needed. It maintains a fixed cache budget for memory usage, enabling efficient read operations.
The writable layer uses a log-structured-merge (LSM) approach, where writes are logged and processed without affecting the read path, ensuring that read queries remain fast and predictable.