There is more consensus in Egalitarian parliaments

Abstract

This paper describes the design and implementation of Egalitarian Paxos (EPaxos), a new distributed consensus algorithm based on Paxos. EPaxos achieves three goals:(1) optimal commit latency in the wide-area when tolerating one and two failures, under realistic conditions; (2) uniform load balancing across all replicas (thus achieving high throughput); and (3) graceful performance degradation when replicas are slow or crash.Egalitarian Paxos is to our knowledge the first protocol to achieve the previously stated goals efficiently-that is, requiring only a simple majority of replicas to be nonfaulty, using a number of messages linear in the number of replicas to choose a command, and committing commands after just one communication round (one round trip) in the common case or after at most two rounds in any case. We prove Egalitarian Paxos's properties theoretically and demonstrate its advantages empirically through an implementation running on Amazon EC2.

Separating the WHEAT from the Chaff: An Empirical Design for Geo-Replicated State Machines

“…During the course of this evaluation, we obtained some unexpected results. The most notorious example is related with the use of multiple leaders -a widely accepted optimization used by several WANoptimized protocols such as Mencius [23] and EPaxos [24]. Specifically, our results indicate that this optimization does not bring significant latency reduction just by itself; instead, we observed that using a fixed leader in a fast replica is a more effective (and simpler) strategy to reduce latency.…”

“…In this section we present the experiments conducted to assess the effectiveness of certain optimizations proposed for SMR in wide area networks [8], [19], [21], [23], [24], [30], [31] and quorum systems [15], [25]. Before presenting our results, we describe some general aspects of our methodology.…”

“…This protocol, which usually requires multiple communication steps, is responsible for a significant latency overhead when SMR is employed for geo-replication. To mitigate this problem, many SMR protocols have been proposed for wide area networks (WANs) (e.g., [1], [23], [24], [30]). These WAN SMR protocols employ optimizations to reduce latency, usually by decreasing the number of communication steps across the WAN.…”

“…WHEAT was designed to operate both under crash and Byzantine faults. To the best of our knowledge, WHEAT is the first SMR protocol that is both optimized for geo-replication and capable of withstanding general Byzantine faults; Mencius [23] and EPaxos [24] tolerate only crash faults while BFT protocols like EBAWA [30] or Steward [1] requires either each replica to have a trusted component that can only fail by crash, or only tolerate Byzantine faults within a site (i.e., do not tolerate compromised sites), respectively.…”

Communication-Closed Layers as Paradigm for Distributed Systems: A Manifesto

“…However, the huge number of involved com-modity computers means that single components can fail very often, and fault tolerance is required as faults become part of the normal operation. Hence we see more and more implementations of fault-tolerant distributed algorithms [48], [30], [36], [5] for data centers.…”