On the Availability of Non-strict Quorum Systems

Aiyer, Amitanand S.; Alvisi, Lorenzo; Bazzi, Rida A.

doi:10.1007/11561927_6

Cited by 37 publications

(43 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Aiyer et al defined k-atomicity, which generalizes Lamport's atomicity by allowing a bounded degree of version-based staleness: each read may return the value assigned by one of the k most recent writes [10]. In contrast, the ∆-atomicity property allows a bounded degree of time-based staleness [8].…”

Section: Background and Related Workmentioning

confidence: 94%

“…The latter point, which we explain in Section III, makes Γ more robust against clock skew. Furthermore, Γ enables more comprehensive benchmarking of eventual consistency than existing work on metrics [10], [11], [12], [13] and measurements [5], [6], [7]. • We apply Γ to perform an offline analysis of Cassandra, a popular open-source key-value store [14].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Client-Centric Benchmarking of Eventual Consistency for Cloud Storage Systems

Golab

Rahman

AuYoung³

et al. 2014

2014 IEEE 34th International Conference on Distributed Computing Systems

View full text Add to dashboard Cite

Eventually-consistent key-value storage systems sacrifice the ACID semantics of conventional databases to achieve superior latency and availability. However, this means that client applications, and hence end-users, can be exposed to stale data. The degree of staleness observed depends on various tuning knobs set by application developers (customers of key-value stores) and system administrators (providers of key-value stores). Both parties must be cognizant of how these tuning knobs affect the consistency observed by client applications in the interest of both providing the best end-user experience and maximizing revenues for storage providers. Quantifying consistency in a meaningful way is a critical step toward both understanding what clients actually observe, and supporting consistency-aware service level agreements (SLAs) in next generation storage systems. This paper proposes a novel consistency metric called Γ (Gamma) that captures client-observed consistency. This metric provides quantitative answers to questions regarding observed consistency anomalies, such as how often they occur and how bad they are when they do occur. We argue that Γ is more useful and accurate than existing metrics. We also apply Γ to benchmark the popular Cassandra key-value store. Our experiments demonstrate that Γ is sensitive to both the workload and client-level tuning knobs, and is preferable to existing techniques which focus on worst-case behavior.

show abstract

Section: Background and Related Workmentioning

confidence: 94%

Section: Introductionmentioning

confidence: 99%

Client-Centric Benchmarking of Eventual Consistency for Cloud Storage Systems

Golab

Rahman

AuYoung³

et al. 2014

2014 IEEE 34th International Conference on Distributed Computing Systems

View full text Add to dashboard Cite

show abstract

“…While we study probabilistic guarantees for staleness, prior work on k-quorums 2,3 have looked at deterministic guarantees that a partial quorum system will return values that are within k versions of the most recent write. 2 Finally, recent research has focused on measuring and verifying the consistency of eventually consistent systems both theoretically and experimentally (Rahman et al provide a brief survey 21 ). This is useful for validating consistency predictions and understanding staleness violations.…”

Section: Wars Limitations and Extensionsmentioning

confidence: 99%

Quantifying eventual consistency with PBS

et al. 2013

View full text Add to dashboard Cite

Data replication results in a fundamental trade-off between operation latency and consistency. At the weak end of the spectrum of possible consistency models is eventual consistency, which provides no limit to the staleness of data returned. However, anecdotally, eventual consistency is often "good enough" for practitioners given its latency and availability benefits. In this work, we explain this phenomenon and demonstrate that, despite their weak guarantees, eventually consistent systems regularly return consistent data while providing lower latency than their strongly consistent counterparts. To quantify the behavior of eventually consistent stores, we introduce Probabilistically Bounded Staleness (PBS), a consistency model that provides expected bounds on data staleness with respect to both versions and wall clock time. We derive a closed-form solution for version-based staleness and model real-time staleness for a large class of quorum replicated, Dynamo-style stores. Using PBS, we measure the trade-off between latency and consistency for partial, non-overlapping quorum systems under Internet production workloads. We quantitatively demonstrate how and why eventually consistent systems frequently return consistent data within tens of milliseconds while offering large latency benefits.

show abstract

“…Existing solutions can be classified into trace-based verifications [7], [9] and benchmark-based verifications [16]- [19]. Trace-based verifications focus on three consistency semantics: safety, regularity, and atomicity, which are proposed by Lamport [20], and extended by Aiyer et al [21]. A register is safe if a read that is not concurrent with any write returns the value of the most recent write, and a read that is concurrent with a write can return any value.…”

mentioning

confidence: 99%

Consistency as a Service: Auditing Cloud Consistency

Liu

Wang

2014

IEEE Trans. Netw. Serv. Manage.

View full text Add to dashboard Cite

Abstract-Cloud storage services have become commercially popular due to their overwhelming advantages. To provide ubiquitous always-on access, a cloud service provider (CSP) maintains multiple replicas for each piece of data on geographically distributed servers. A key problem of using the replication technique in clouds is that it is very expensive to achieve strong consistency on a worldwide scale. In this paper, we first present a novel consistency as a service (CaaS) model, which consists of a large data cloud and multiple small audit clouds. In the CaaS model, a data cloud is maintained by a CSP, and a group of users that constitute an audit cloud can verify whether the data cloud provides the promised level of consistency or not. We propose a two-level auditing architecture, which only requires a loosely synchronized clock in the audit cloud. Then, we design algorithms to quantify the severity of violations with two metrics: the commonality of violations, and the staleness of the value of a read. Finally, we devise a heuristic auditing strategy (HAS) to reveal as many violations as possible. Extensive experiments were performed using a combination of simulations and real cloud deployments to validate HAS.Index Terms-Cloud storage, consistency as a service (CaaS), two-level auditing, heuristic auditing strategy (HAS).

show abstract

On the Availability of Non-strict Quorum Systems

Cited by 37 publications

References 26 publications

Client-Centric Benchmarking of Eventual Consistency for Cloud Storage Systems

Client-Centric Benchmarking of Eventual Consistency for Cloud Storage Systems

Quantifying eventual consistency with PBS

Consistency as a Service: Auditing Cloud Consistency

Contact Info

Product

Resources

About