Contention-related crash failures: Definitions, agreement algorithms, and impossibility results

This article explores an interplay between process crash failures and concurrency. Namely, it aims at answering the question, “Is possible to cope with more when some number of crashes occur before predefined contention point happened?”. These are named λ-constrained crashes, where λ is (known by processes). Hence, this considers two types crashes: classical (which can any time consequently called any-time crashes). Considering a system made up n asynchronous processes communicating through atomic read/write registers, focuses on design agreement-related algorithms. Assuming λ=n−1 no failure, first algorithm solves consensus problem in presence one thereby circumventing well-known FLP impossibility result. The second k-set agreement for k≥2. It such that λ=n−ℓ ℓ≥k=m+f works (2m+ℓ−k) (f−1) i.e., t=(2m+ℓ−k)+(f−1) crashes. follows considering timing respect enlarges space executions which be solved despite combined effect asynchrony, concurrency, paper also presents results context (with or without