Consensus is arguably the most studied problem in distributed computing as a whole, and particularly in the distributed message-passing setting. In this latter framework, research on consensus has considered various hypotheses regarding the failure types, the memory constraints, the algorithmic performances (e.g., early stopping and obliviousness), etc. Surprisingly, almost all of this work assumes that messages are passed in a complete network, i.e., each process has a direct link to every other process. A noticeable exception is the recent work of Castañeda et al. (Inf. Comput. 2023) who designed a generic oblivious algorithm for consensus running in radius(G,t) rounds in every graph G, when up to t nodes can crash by irrevocably stopping, where t is smaller than the node-connectivity κ of G. Here, radius(G,t) denotes a graph parameter called the radius of G whenever up to t nodes can crash. For t = 0, this parameter coincides with radius(G), the standard radius of a graph, and, for G = K_n, the running time radius(K_n,t) = t+1 of the algorithm exactly matches the known round-complexity of consensus in the clique K_n. Our main result is a proof that radius(G,t) rounds are necessary for oblivious algorithms solving consensus in G when up to t nodes can crash, thus validating a conjecture of Castañeda et al., and demonstrating that their consensus algorithm is optimal for any graph G. We also extend the result of Castañeda et al. to two different settings: First, to the case where the number t of failures is not necessarily smaller than the connectivity κ of the considered graph; Second, to the k-set agreement problem for which agreement is not restricted to be on a single value as in consensus, but on up to k different values.

We develop a new lower bound for k-set agreement in synchronous message-passing systems connected by an arbitrary directed communication network, where up to t processes may crash. Our result thus generalizes the ⌊t/k⌋ + 1 lower bound for complete networks in the t-resilient model by Chaudhuri, Herlihy, Lynch, and Tuttle [JACM 2000]. Moreover, it generalizes two lower bounds for oblivious algorithms in synchronous systems connected by an arbitrary undirected communication network known to the processes, namely, the domination number-based lower bound by Castañeda, Fraigniaud, Paz, Rajsbaum, Roy, and Travers [TCS 2021] for failure-free processes, and the radius-based lower bound in the t-resilient model by Fraigniaud, Nguyen, and Paz [STACS 2024]. Our topological proof non-trivially generalizes and extends the connectivity-based approach for the complete network, as presented in the book by Herlihy, Kozlov, and Rajsbaum (2013). It is based on a sequence of shellable carrier maps that, starting from a shellable input complex, determine the evolution of the protocol complex: During the first ⌊t/k⌋ rounds, carrier maps that crash exactly k processes per round are used, which ensure high connectivity of their images. A Sperner’s lemma style argument can thus be used to prove that k-set agreement is still impossible by that round. From round ⌊t/k⌋ + 1 up to our actual lower bound, a novel carrier map is employed, which maintains high connectivity. As a by-product, our proof also provides a strikingly simple lower-bound for k-set agreement in synchronous systems with an arbitrary communication network, where exactly t ≥ 0 processes crash initially, i.e., before taking any step. We demonstrate that the resulting additional agreement overhead can be expressed via an appropriately defined radius of the communication graphs, and show that the usual input pseudosphere complex for k-set agreement can be replaced by an exponentially smaller input complex based on Kuhn triangulations, which we prove to be also shellable.

International audience

Consensus is arguably the most studied problem in distributed computing as a whole, and particularly in distributed message-passing settings. Research on consensus has considered various failure types, memory constraints, and much more. Surprisingly, almost all of this work assumes that messages are passed in a complete network, i.e., each process has a direct link to every other process. Set agreement, a relaxed variant of consensus, has also been heavily studied in different settings, yet research on it has also been limited to complete networks. We address this situation by considering consensus and set agreement in general networks, i.e., that can have an arbitrary graph G as their communication graph. We focus on fault-prone networks, where up to t nodes may crash and irrevocably stop communicating, and present upper and lower bounds for such networks. We establish the following collection of results: - The consensus algorithm by [Castañeda et al., 2023] is optimal for all graphs, and not only for symmetric graphs. - This algorithm can be extended to a generic algorithm for k-set agreement, for every k ≥ 1. For k = 1, our generic algorithm coincides with the existing one for consensus. - All these algorithms can be extended to the case where the number t of failures exceeds the connectivity κ of the graph, while the existing consensus algorithm assumed that t < κ.