Best Paper Award
A Novel Detection and Recovery Techniques for Physical Defect in FPGA-IP Cores
Motoki Amagasaki, Yuki Nishitani, Kazuki Inoue, Masahiro Iida, Morihiro Kuga,
Toshinori Sueyoshi
[Trans. Inf. & Syst. (JPN Edition), Vol. J96-D No.12, Dec. 2013]

Motoki Amagasaki

Yuki Nishitani

Kazuki Inoue

Masahiro Iida

Morihiro Kuga

Toshinori Sueyoshi
   
  Fault-tolerant systems play an important role in today's advanced information-intensive society. Current fault-tolerant systems require not only high tolerance to faults, but also low overhead, e.g., small area, low power, and short delay. In this paper, the authors propose a fault-tolerant system based on FPGA (field-programmable gate array) - IP (intellectual property) cores. An FPGA is a flexible device -- it can be arbitrarily configured in the field. The proposed system utilizes this flexibility of FPGAs for fault tolerance. When detecting a failure, this system can detach the faulty parts responsible for this failure and reconfigure itself with the remaining fault-free parts, and thus recover from the failure. Note that, unlike N-Modular Redundancy, this system does not require any redundancy, so that it can keep the overhead low.
  There are two important issues for this kind of fault-tolerant systems fault detection (or identification) and reconfiguration. The authorsf novel fault detection method, which is an extension of their previous production testing method, can identify the faulty part completely with a short testing time. Their reconfiguration method, on the other hand, can efficiently recovery from failure. This is because it employs a CAD (computer-aided design) tool, which has high placing and routing ability. Experimental results show that the proposed system can completely identify given faults and avoid them with almost no performance degradation; it can be applied to practical systems requiring high reliability.
  Since FPGA-IP cores are expected to be frequently and widely utilized in the future, fault-tolerant systems with FPGA-IP cores will be a mainstream approach to obtaining high-reliability systems. This paper, in which such a fault-tolerant system is clearly designated, is worthy of the IEICE Best Paper Award.

Close