Brief Announcement: On the Meaning of Solving a Task with a Failure Detector
We amend the framework, of two decades, of failure detectors [3,4] to bring it in line with the modern view of solving a distributed task  that separates processes and threads. While the conventional framework precludes a thread from advancing in the absence of failure detector values to “its” process, we allow live processes to advance the threads of failed processes. This provides for the application of the wealth of simulation techniques [2,6,7] designed for read-write threads and consequently to completely characterize task solvability with failure detectors. When dealing with the extremes, consensus and set-consensus, the former framework sufficed. With the advances in understanding of more nuanced notions like k-set consensus the framework requires amendment.
- 1.Afek, Y., Nir, I.: Failure detectors in loosely named systems. In: PODC, pp. 65–74. ACM Press, New York (2008)Google Scholar
- 2.Borowsky, E., Gafni, E.: Generalized FLP impossibility result for t-resilient asynchronous computations. In: STOC, pp. 91–100. ACM Press, New York (1993)Google Scholar
- 5.Delporte-Gallet, C., Fauconnier, H., Guerraoui, R.: Tight failure detection bounds on atomic object implementations. J. ACM 57(4) (2010)Google Scholar