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
Lobsters·15d ago·5 min read
Who needs a real-time database?
TL;DR
Real-time databases (RDBs) are often seen as essential in kdb+ systems, but they may not always be necessary. By comparing in-memory and historical databases (HDBs), the author argues that HDBs can perform comparably when the OS page cache is warm.
✦ Why It Matters
Engineers should assess whether to use RDBs or HDBs based on their specific application needs and data access patterns.
Key Takeaways
- Evaluate if rdbs are necessary for your kdb+ applications.
- Leverage hdbs for faster concurrent query execution.
- Consider simplifying your architecture by removing rdbs.
- Utilize the OS page cache to improve data access speeds.
How It Works
The proposed architecture uses hdbs to store data on disk while leveraging the OS page cache for fast access. This allows multiple queries to run concurrently, improving performance compared to the serial processing of rdbs.