A High-Level Approach to Analyze the Effects of Soft Errors on Lossless Compression Algorithms

详细信息    查看全文

• 作者：Serhiy Avramenko ; Matteo Sonza Reorda ; Massimo Violante…
• 关键词：Soft errors ; Lossless compression ; Reliability ; High ; level fault injection
• 刊名：Journal of Electronic Testing
• 出版年：2017
• 出版时间：February 2017
• 年：2017
• 卷：33
• 期：1
• 页码：53-64
• 全文大小：
• 刊物类别：Engineering
• 刊物主题：Circuits and Systems; Electrical Engineering; Computer-Aided Engineering (CAD, CAE) and Design;
• 出版者：Springer US
• ISSN：1573-0727
• 卷排序：33

文摘

In space applications, the data logging sub-system often requires compression to cope with large amounts of data as well as with limited storage and communication capabilities. The usage of Commercial off-the-Shelf (COTS) hardware components is becoming more common, since they are particularly suitable to meet high performance requirements and also to cut the cost with respect to space qualified ones. On the other side, given the characteristics of the space environment, the usage of COTS components makes radiation-induced soft errors highly probable. The purpose of this work is to analyze a set of lossless compression algorithms in order to compare their robustness against soft errors. The proposed approach works on the unhardened version of the programs, aiming to estimate their intrinsic robustness. The main contribution of the work lies in the investigation of the possibility of performing an early comparison between different compression algorithms at a high level, by only considering their data structures (corresponding to program variables). This approach is virtually agnostic of the downstream implementation details. This means that the proposed approach aims to perform a comparison (in terms of robustness against soft errors) between the considered programs before the final computing platform is defined. The results of the high-level analysis can also be used to collect useful information to optimize the hardening phase. Experimental results based on the OpenRISC processor are reported. They suggest that when properly adopted, the proposed approach makes it possible to perform a comparison between a set of compression algorithms, even with a very limited knowledge of the target computing system.