Faster or better designed? Choose any two! by Dmytro Vyazelenko

Rust

Explore the trade-offs between performance and design principles in modern systems. Discover how to achieve low latency, reduce memory allocation and improve system efficiency, and learn from real-world examples.

Key takeaways

No garbage allocation is a key principle in achieving low latency and reducing allocations in systems.
C and Rust based implementations like Agrona can outperform Java-based implementations in certain scenarios.
Java’s garbage collector and object-oriented programming model can lead to increased latency and memory allocation.
Using custom, low-level, buffer-based APIs can help reduce allocation and improve performance.
Avoid using text-oriented protocols like JSON and XML, as they can introduce additional overhead and latency.
Batching and concurrency can be used to improve performance, but require careful consideration of thread concurrency and resource contention.
On-premises solutions can provide better performance and lower latency than cloud-based solutions, but require careful planning and optimization.
Design principles like no garbage allocation, batching, and concurrency can help achieve low latency and high performance in systems.

Faster or better designed? Choose any two! by Dmytro Vyazelenko

More talks