EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Aug 18·5 min read
Why I reimplemented LVM (with worse guarantees)
TL;DR
A new implementation of LVM was created to optimize microVM launches on bare-metal hypervisors, prioritizing speed and simplicity over traditional guarantees. This approach significantly reduces boot time, enhancing performance for customer workloads.
✦ Why It Matters
Consider simplifying your storage management tools to meet specific workload demands for faster deployment.
Key Takeaways
- Evaluate if existing tools meet your performance needs.
- Consider custom solutions for specific workload requirements.
- Utilize Linux kernel's device-mapper for efficient storage management.
- Prioritize speed over safety when data loss is acceptable.
How It Works
The new implementation uses a custom memory allocator that operates in-process, allowing for rapid allocation of block devices without the delays associated with LVM's global locking. By managing memory directly and leveraging the kernel's asynchronous device-mapper, the system achieves significantly faster performance, crucial for launching microVMs quickly.