Using Real Time Linux Kernel with Computationally Intensive Workloads

Abstract

This article is devoted to the problems of running in the cloud computing environment scientific and engineering applications, which require both a large amount of calculations and also a precise temporary resolution of operations. Such requirements typically arise in applications that receive data from the hardware, process it, and the results of processing are used to issue control commands for the equipment. The article considers the possibility of using a real-time Linux kernel to support this class of applications. Different hardware was tested, the results were obtained for achievable latency for real-time processes, and also the effect of the real-time kernel on the performance of the computing subsystem was measured (the number of GFlops in LINPACK tests, and the speed of working with memory on NUMA systems).

Key words: real time linux, soft real time, preempt_rt