Software-controlled fault tolerance

Abstract

Traditional fault tolerance techniques typically utilize resources ineffectively because they cannot adapt to the changing reliability and performance demands of a system. This paper proposes software-controlled fault tolerance, a concept allowing designers and users to tailor their performance and reliability for each situation. Several software-controllable fault detection techniques are then presented: SWIFT, a software-only technique, and CRAFT, a suite of hybrid hardware/ software techniques. Finally, the paper introduces PROFiT, a technique which adjusts the level of protection and performance at fine granularities through software control. When coupled with software-controllable techniques like SWIFT and CRAFT, PROFiT offers attractive and novel reliability options.

Assuring application-level correctness against soft errors

“…Hence, whenever any edge weight crosses the threshold value, we switch to version V 1 as we can no longer guarantee non-criticality of instructions in version V 2 due to the unexpected dependence. Note that since [20] is a software-only based fault tolerance scheme, there are certain kinds of hardware errors which cannot be detected by such a scheme -for more details, we refer the reader to [34]. While we choose to use the scheme in [20] for our experiments, it is important to note that any improvements to software-only fault tolerant schemes (such as hardware support for correct update of the PC) are orthogonal to our technique and hence can be jointly used.…”

Predicting robustness against transient faults of MPI based programs

“…On Software-Controlled Fault Tolerance [8], the authors presented a set of software and hybrid (software and hardware) transient fault detection techniques. Each of the proposed techniques had a different cost/benefit relation by improving reliability or performance.…”

“…Continuing their research in fault tolerance for transient faults, the same authors of [8] proposed Spot [9], a technique to dynamically insert redundant instructions to detect errors generated by transient faults. This dynamic insertion was made in runtime using instrumentation.…”

Efficient soft error protection for commodity embedded microprocessors using profile information

“…Other works such as CRAFT and PROFIT [27] improve upon the SWIFT solution by leveraging additional hardware structures and architectural vulnerability factor (AVF) analysis [23], respectively. Compiler-based instruction duplication delivers nearly complete fault coverage, with the added benefit of requiring little to no hardware cost.…”

“…Although symptom-based detection is inexpensive, the amount of coverage that can be obtained from a symptom-only approach is typically limited. To address this limitation, we make use of the second area of prior research, software-based instruction duplication [26,27]. With this approach, instructions are duplicated and results are validated within a single thread of execution.…”