Eventually Perfect Failure Detectors Using ADD Channels

Abstract

Abstract. We present a novel implementation of the eventually perfect failure detector (3P) from the original hierarchy of Chandra-Toueg oracles. Previous implementations of 3P have assumed models of partial synchrony where point-to-point message delay is bounded and/or communication is reliable. We show how to implement this important oracle under even weaker assumptions using Average Delayed/Dropped (ADD) channels. Briefly, all messages sent on an ADD channel are privileged or non-privileged. All non-privileged messages can be arbitrarily delayed or even dropped. For each run, however, there exists an unknown window size w, and two unknown upper-bounds d and r, where d bounds the average delay of the last w privileged messages, and r bounds the ratio of non-privileged messages to privileged messages per window.

Crash fault detection in celerating environments

“…Although we have demonstrated the use of bichronal timers only in a ping-ack based 3P implementation, bichronal timers may be used in heartbeat-based 3P implementations too (e.g. [2,23,5,4]). In heartbeat-based implementations, the local failure detector module at each process p sends periodic "heartbeats" to all the processes that monitor p. The monitoring processes then adaptively estimate the inter-arrival time of these heartbeats 7 .…”

“…Recently proposed models like the Finite Average Response Time (FAR) model [11] and the Average Delay/Drop (ADD) model [23] make additional assumptions on absolute process speeds to circumvent the celeration problem. For instance, the FAR model [11] assumes an upper bound on absolute process speeds 3 .…”

“…This assumption precludes continual acceleration of processes and allows failure detector implementations in the FAR model to use action clocks which are immune to continual process deceleration. In contrast, the ADD model [23] assumes a lower bound on absolute process speeds. This assumption precludes continual deceleration of processes and allows failure detector implementations in the ADD model to use real-time clocks which are immune to continual process acceleration.…”

Failure Detectors Encapsulate Fairness

“…Despite such uncertainty, partial synchrony is useful for solving problems in crash-prone distributed systems, and several such models have been proposed in the literature (e.g., [18,17,26,27,[39][40][41]). These models vary in the information they provide about these bounds, and consequently they have different crash detection capabilities.…”

Wait-Free Dining Under Eventual Weak Exclusion

“…There are known implementations of 3P in several other models partial synchrony as well [3,[16][17][18]. The common advantage is that 3P-based algorithms are decoupled from explicit commitments to underlying detection mechanisms and/or specific timing parameters.…”