Task replication and control for highly parallel in-memory stores
Fernando Martin Del Campo, Paul Chow · Proceedings of the International Symposium on Memory Systems · 2017
The computing world has made significant investments and developments around the von Neumann architecture. This means that applications have been highly tuned to the CPU-based server, which may not be the best platform for some applications. In this paper, we examine a class of large-scale, data-intensive applications by first abstracting away the architectural details to the essential operations. By doing this, we can study the intrinsic behavior and requirements of these applications and then find the hardware architecture that can implement the required operations most efficiently.