Enabling Low Tail Latency on Multicore Key-Value Stores
We present RStore to enable low and predictable latency (i.e. low tail latency) and efficient use of hardware resources such as CPU, memory and storage through the following design points:
- Asynchronous execution
- Hybrid DRAM+NVM architecture
- Log-structured storage
- User-space networking
Enabling Low Tail Latency on Multicore Key-Value Stores
https://tong1heng.github.io/2022/04/05/Embedded/p1091-lersch/