We consider the anonymous broadcast model: a set of n anonymous processes communicate via send-to-all primitives. We assume that underlying communication channels are asynchronous but reliable, and that the processes are subject to crash failures. We show first that in this model, even a single faulty process precludes implementations of atomic objects with non-commuting operations, even as simple as read-write registers or add-only sets. We, however, show that a sequentially consistent read-write memory and add-only sets can be implemented t-resiliently for t<n/2, i.e., provided that a majority of the processes do not fail. We use this implementation to establish an equivalence between the t-resilient read-write anonymous shared-memory model and the t-resilient anonymous broadcast model in terms of colorless task solvability. As a result, we obtain the first task computability characterization for unreliable anonymous message-passing systems.
@InProceedings{kuznetsov_et_al:LIPIcs.OPODIS.2018.23, author = {Kuznetsov, Petr and Yanagisawa, Nayuta}, title = {{Task Computability in Unreliable Anonymous Networks}}, booktitle = {22nd International Conference on Principles of Distributed Systems (OPODIS 2018)}, pages = {23:1--23:13}, series = {Leibniz International Proceedings in Informatics (LIPIcs)}, ISBN = {978-3-95977-098-9}, ISSN = {1868-8969}, year = {2019}, volume = {125}, editor = {Cao, Jiannong and Ellen, Faith and Rodrigues, Luis and Ferreira, Bernardo}, publisher = {Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik}, address = {Dagstuhl, Germany}, URL = {https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.OPODIS.2018.23}, URN = {urn:nbn:de:0030-drops-100830}, doi = {10.4230/LIPIcs.OPODIS.2018.23}, annote = {Keywords: Distributed tasks, anonymous broadcast, fault-tolerance} }
Feedback for Dagstuhl Publishing