How Application-Level Priority Management Keeps Latency Low and Throughput High
Throughput and latency are at a constant tension. Optimizing for throughput requires running machines at high utilization, which increases queuing delays and hurts latency. This Linux Foundation talk by ScyllaDB CTO Avi Kivity shows how high throughput and low latency can both be achieved in a single application by using application-level priority scheduling.
- Vertical scalability, or how a single node is a network
- Shard-per-core, shared nothing design
- Async everywhere
- Schedulers for CPU and IO
- Prioritizing multiple workloads on the same cluster
- Getting the most out of the Linux OS and C++20