What Modern NVMe Storage Can Do, And How To Exploit It: High-Performance I/O for High-Performance Storage Engines

CacheSack: Admission Optimization for Google Datacenter Flash Caches

This paper introduces CacheSack, a decentralized admission algorithm designed to optimize Google’s datacenter flash caches by minimizing the total cost of ownership. The system effectively reduces expenses by balancing the trade-off between disk I/O performance and flash wearout caused by limited write endurance. By partitioning traffic into categories and modeling cache behavior as a knapsack problem, CacheSack dynamically selects the most efficient admission policy for each workload in real time. Production results demonstrate that this automated approach improves total operational costs by 6.5% compared to previous manual methods. It further achieves a significant 26% reduction in flash writes while simultaneously decreasing disk reads. Ultimately, CacheSack provides a scalable, low-overhead solution that adapts to diverse database workloads without requiring human intervention.

Google Colossus: Scaling Distributed Storage and Technical Quality Engineering

Benchmarking NoSQL Key-Value Stores: Redis, Aerospike, and Dragonfly Performance

Forensic Analysis of Redis Databases via RDB and AOF Files

Scalable and Reliable In-Memory Storage: A Redis Case Study

Redis: Optimized Memory Management and Cache-Friendly Indexing

Performance Benchmarking of Redis and SSDB NoSQL Databases

Scalable Causal Consistency with COPS

Scaling Replicated State Machines with Compartmentalization