Building lock-free data structures

In this section, we will discuss data structures that are shared between threads or between processes in an HFT ecosystem and the concurrency and synchronization considerations involved, especially given the extremely high-throughput and low-latency requirements of HFT systems. We will also design and discuss the performance implications of lock-free data structures. A lock-free data structure is a mechanism to share data between producers and consumers running in different threads and/or processes. The interesting point here is that it achieves this by avoiding locks altogether and thus performs significantly better for HFT purposes. We will investigate these items in more detail in the sections that follow.

When/why are locks needed (non-HFT applications)

Let's look at why locks are needed in a traditional multi-threaded/multi-process programming paradigm. The reason comes down fundamentally to the need to allow concurrent access to...