Towards implementing multi-layer reflection for fault-tolerance
François Taı̈ani, J. Fabre, Marc‐Olivier Killijian · 2004
Th3r2 party software is now in reasingly used in systems with hhm dependability requirements. Thq evolution of system development raises new h czM55LcO in parti ular regarding thg implementation of faulttoleran e. As systems are often built of bla k-box omponents, some ru ial aspe ts of thczz behzz5 regarding repli ation annot be h cWM ThW is also true to some extent for open-sour e omponents as mastering thste internal behnal c is sometimes very tri ky (e.g. OS and ORBs). During thi last de ade refle tion ho emerged as a very fruitful paradigm for dis iplined management of non-fun tional aspe ts, among wh h fault-toleran e. In thcW paper we dis uss hs to apply refle tion to multi-layer systems for implementing faulttoleran e in an independent and prin ipled manner. We analyze thl onne tions between thw underlying assumptions of fault-toleran e strategies and different layers of a system. Based on thcW multi-layer analysis we shcz hc thz requirements of a family of repli ation algorithz an be addressed on a on rete arhcz ture, resulting in whWLchMWLchchhO5Lzchzc tion.