Research of Fault-Tolerant Computing Using COTS Elements
B. Swiercz, D. Makowski and A. Napieralski
Technical University of Lodz, PL
single event upset, radiation environment, fault tolerant system, hardened, real-time system
The problem of designing hardened systems is very important for modern physics. The physics experiences design in linear accelerators and synchrotrons need measurements and control devices, which are under radiation impact. This paper highlights the operating system, which is used to develop a software fault tolerant environment for reliable computing. The main idea is to build fault tolerant algorithms into kernel instead of user-space’s programs.
To design a new European research project, X-Ray Free Electron Laser (X-FEL), sophisticated distributed control systems must be used. Many of those devices (especially embedded computers) work close to cavities that produce high gamma and neutrons. High energy particles (e.g. neutrons) can affect memory and registers in microprocessor systems and cause on illegal operation and thus loss of the system functionality. The fault model is the single transient bit-flip, which is known as a Single Event Upset (SEU). Therefore, it is recommended to design redundancy systems not only on the hardware layer but also on the software one. Implementation of redundancy in the software instead of hardware should reduce designing and testing cost of the main control system. The problem is that modern kernels, like Linux kernel, are to complicated for using to research fault tolerant algorithms. In this papers new real-time kernel, called sCore, is proposed.
Back to Program
Nanotech 2006 Conference Program Abstract