How do you design cloud systems for low-latency applications

Question

I'm building a cloud-based system where low latency is critical. What design patterns or services should I consider to keep latency as low as possible across the stack?