Probabilistic Fault-Containment

Fast Leader (Full) Recovery Despite Dynamic Faults

“…Correcting the global state quickly was investigated in the context of self-stabilization for the purpose of fault containment [14,15,2,3,8] (that is, preventing local memory corruptions from propagating to the whole network). The state of the art in this matter nevertheless requires that either only a single process is corrupted [14,3], faulty processes are surrounded by many correct ones so that few faults can be caught quickly [3], the network is fully synchronous [15], or the recovery guarantee is only probabilistic [8].…”

“…The state of the art in this matter nevertheless requires that either only a single process is corrupted [14,3], faulty processes are surrounded by many correct ones so that few faults can be caught quickly [3], the network is fully synchronous [15], or the recovery guarantee is only probabilistic [8]. The "fault gap" that results from those approaches is significantly reduced, as only a delay that depends on the fault span must separate consecutive faults.…”

“…With respect to "output stabilization" [4,5,11,12,1,6,7,16,17,1,6], our protocol recovers the full correct state quickly (in time O(k 2 )). With respect to fault-containment [14,15,3,8], LE(k) can handle up to k faults [13,14,3], faults can be arbitrarily spread and the network is sparse [3], the network is fully asynchronous and the scheduling is unfair [15], and the recovery property is deterministic [8].…”

“…p.R_probe and p.R_report are either ⊥ or non-positive. 8. p.num_L_null, p.num_R_null, non-negative integers, in the range {0, .…”

Fault-Containing Self-Stabilization in Asynchronous Systems with Constant Fault-Gap

“…They apply the technique to the persistent-bit problem [4] and leader election [5]. The technique seems to be suitable for weakly-stabilizing systems [5] only.…”