Latency-Aware Leader Selection for Geo-Replicated Byzantine Fault-Tolerant Systems

Eischer, Michael; Distler, Tobias

doi:10.1109/dsn-w.2018.00053

Cited by 17 publications

(23 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note that this is not only true with regard to the task of agreeing on requests during normal operation, but for example also for electing a new leader as part of fault handling. (2) Due to the fact that all requests must flow through the leader, the geographic location of the leader, and in particular its position relative to the majority of followers, usually has a significant influence on latency [23,46]. Consequently, a leader switch may decisively change a system's performance characteristics, requiring clients to deal with the associated latency volatility.…”

Section: Existing Approachesmentioning

confidence: 99%

Resilient Cloud-based Replication with Low Latency

Eischer

Distler

2020

Proceedings of the 21st International Middleware Conference

Self Cite

View full text Add to dashboard Cite

Existing approaches to tolerate Byzantine faults in geo-replicated environments require systems to execute complex agreement protocols over wide-area links and consequently are often associated with high response times. In this paper we address this problem with Spider, a resilient replication architecture for geo-distributed systems that leverages the availability characteristics of today's public-cloud infrastructures to minimize complexity and reduce latency. Spider models a system as a collection of loosely coupled replica groups whose members are hosted in different cloudprovided fault domains (i.e., availability zones) of the same geographic region. This structural organization makes it possible to achieve low response times by placing replica groups in close proximity to clients while still enabling the replicas of a group to interact over short-distance links. To handle the inter-group communication necessary for strong consistency Spider uses a reliable group-to-group message channel with first-in-first-out semantics and built-in flow control that significantly simplifies system design.CCS Concepts: • Computer systems organization → Dependable and fault-tolerant systems and networks.

show abstract

Section: Existing Approachesmentioning

confidence: 99%

Resilient Cloud-based Replication with Low Latency

Eischer

Distler

2020

Proceedings of the 21st International Middleware Conference

Self Cite

View full text Add to dashboard Cite

show abstract

“…A variety of research touches the fields of SMR optimizations in WAN environments [13], [14], [17]- [19] and dynamic approaches for latency awareness [16], [20]- [22].…”

Section: Related Workmentioning

confidence: 99%

“…Further research work shows that clients can also dynamically react to changing workloads by efficiently changing their quorum selections to achieve good performance [22]. A protocol for latency-aware leader selection in geo-replicated systems is ARCHER [20], which uses clients' observed end-to-end response latencies to select the optimal leader and hence can dynamically adjust to varying workloads. In contrast, AWARE measures replica-to-replica latencies and uses weight tuning additional to leader selection.…”

Section: Related Workmentioning

confidence: 99%

Resilient Wide-Area Byzantine Consensus Using Adaptive Weighted Replication

Berger

Reiser

Sousa

et al. 2019

2019 38th Symposium on Reliable Distributed Systems (SRDS)

View full text Add to dashboard Cite

In geo-replicated systems, the heterogeneous latencies of connections between replicas limit the system's ability to achieve consensus fast. State machine replication (SMR) protocols can be refined for their deployment in wide-area networks by using a weighting scheme for active replication that employs additional replicas and assigns higher voting power to faster replicas. Utilizing more variability in quorum formation allows replicas to swiftly proceed to subsequent protocol stages, thus decreasing consensus latency. However, if network conditions vary during the system's lifespan or faults occur, the system needs a solution to autonomously adjust to new conditions. We incorporate the idea of self-optimization into geographically distributed, weighted replication by introducing AWARE, an automated and dynamic voting weight tuning and leader positioning scheme. AWARE measures replica-to-replica latencies and uses a prediction model, thriving to minimize the system's consensus latency. In experiments using different Amazon EC2 regions, AWARE dynamically optimizes consensus latency by selfreliantly finding a fast weight configuration yielding latency gains observed by clients located across the globe.

show abstract

“…In [10], Liu and Vukolić proposed two methods for geographic SMR: Droppy that dynamically relocates a set of replication leaders according to given replication settings and workload situations, and Dripple that divides the replicated system state into multiple partitions so that Droppy can efficiently relocate the leaders. Eischer and Distler proposed Archer [11] that relocates leaders based on their response times as measured by clients. A Hash-chain-based technique was employed in the protocol to allow clients to detect illegal phases caused by Byzantine replicas to prevent such replicas from being wrongly assigned as leaders.…”

Section: Related Workmentioning

confidence: 99%

Evaluation and Ranking of Replica Deployments in Geographic State Machine Replication

Shota

Nakamura

Ohmura

2019

2019 38th International Symposium on Reliable Distributed Systems Workshops (SRDSW)

View full text Add to dashboard Cite

Geographic state machine replication (SMR) is a replication method in which replicas of a service are located on multiple continents to improve the fault tolerance of a general service. Nowadays, geographic SMR is easily realized using public cloud services; SMR provides extraordinary resilience against catastrophic disasters. Previous studies have revealed that the geographic distribution of the replicas has a significant influence on the performance of the geographic SMR; however, the optimal way for a system integrator to deploy replicas remains unknown. In this paper, we propose a method to evaluate and rank replica deployments to assist a system integrator in deciding a final replica deployment. In the method, we also propose a novel evaluation function that estimates a latency of SMR protocols with round-trip time (RTT). To demonstrate the effectiveness of the proposed method, we build thousands of geographic SMRs on Amazon Web Services and present experimental results. The results show that the proposed method that estimates a latency based on RTTs can generate consistent rankings with reasonable calculation time.

show abstract

Latency-Aware Leader Selection for Geo-Replicated Byzantine Fault-Tolerant Systems

Cited by 17 publications

References 13 publications

Resilient Cloud-based Replication with Low Latency

Resilient Cloud-based Replication with Low Latency

Resilient Wide-Area Byzantine Consensus Using Adaptive Weighted Replication

Evaluation and Ranking of Replica Deployments in Geographic State Machine Replication

Contact Info

Product

Resources

About