@inproceedings{GK-FDTB-03,
Title = {On Failure Detectors and Type Boosters},
Author = {Guerraoui, Rachid and Kouznetsov, Petr},
Booktitle = {17th International Conference on Distributed Computing (DISC)},
Pages = {292-305},
Year = {2003},
Isbn = {978-3-540-20184-7},
Issn = {0302-9743},
Doi = {http://dx.doi.org/10.1007/978-3-540-39989-6_21},
Location = {Sorrento, Italy},
Address = {Berlin / Heidelberg, Germany},
Volume = {2848},
Publisher = {Springer},
Series = {Lecture Notes in Computer Science (LNCS)},
Abstract = {The power of a set S of object types can be measured as the maximum number n of processes that can solve consensus using only types in S and registers. This number, denoted by h^r_m(S), is called the consensus power of S. The use of failure detectors can however ``boost'' the consensus power of types. This paper addresses the weakest failure detector type booster question, which consists in determining the weakest failure detector D such that, for any set S of types with h_m^r(S)=n, h_m^r(S;D)=n+1. We consider the failure detector Ω_n (introduced in (Neiger, 1995)) which outputs, at each process, a set of at most n processes so that, eventually, all correct processes detect the same set of processes that includes at least one correct process. We prove that Ω_n is the weakest failure detector type booster for deterministic one-shot types. As an interesting corollary of our result, we show that Ω_f is the weakest failure detector to boost the power of (f-1)-resilient objects solving consensus.},
Categories = {tlabs_no},
Projectname = {distributed_systems and thisisimportant}
}