A General Reliability Model for Data Storage Systems

Venkatesan, Vinodh; Iliadis, Ilias

doi:10.1109/qest.2012.32

Cited by 24 publications

(21 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following a first-disk failure, a complex sequence of rebuilds and subsequent disk failures may occur which eventually lead the system either to irrecoverable data loss with probability P DL or back to the original fully-operational mode by restoring all replicas with probability 1 − P DL . In [17] it is shown that, for all failure distributions considered in this article, it holds that…”

Section: System Reliabilitymentioning

confidence: 92%

“…In the area of reliability analysis, a theoretical framework based on the direct path approach has been proposed to study the effect of data placement, redundancy schemes, latent sector errors, failure and rebuild time distributions, and network bandwidth on the reliability of storage systems [8], [9], [17], [18]. In particular, it has been shown that limited network bandwidth can significantly affect rebuild processes for certain data placement schemes [9].…”

Section: Reliability Analysismentioning

confidence: 99%

“…Consequently, our model is also valid in the case when disk failures are correlated within sites. Device Lifetime Distribution: In the analysis of this article, the lifetime distribution of a disk is assumed to be part of a large class of distributions described in [17,Section 3]. This class of distributions is important as it contains real-world disk lifetime distributions, such as Weibull and gamma, and the MTTDL has been shown to be invariant within this class [17].…”

Section: Failure Modelmentioning

confidence: 99%

“…Device Lifetime Distribution: In the analysis of this article, the lifetime distribution of a disk is assumed to be part of a large class of distributions described in [17,Section 3]. This class of distributions is important as it contains real-world disk lifetime distributions, such as Weibull and gamma, and the MTTDL has been shown to be invariant within this class [17]. The intuition behind this is that in a large population of disks, the age distribution of the disks tends to stabilize over time and the average failure rate in the system becomes constant in the steady-state, such that both Weibull and exponential distributions provide the same quality of fit [4].…”

Section: Failure Modelmentioning

confidence: 99%

“…To assess the reliability measures of interest we adopt a methodology analogous to the one presented in [17]. Exposure Levels.…”

Section: Appendixmentioning

confidence: 99%

See 4 more Smart Citations

Reliability of Geo-replicated Cloud Storage Systems

Iliadis

Sotnikov

Ta-Shma

et al. 2014

2014 IEEE 20th Pacific Rim International Symposium on Dependable Computing

Self Cite

View full text Add to dashboard Cite

Network bandwidth between sites is typically more scarce than bandwidth within a site in geo-replicated cloud storage systems, and can potentially be a bottleneck for recovery operations. We study the reliability of geo-replicated cloud storage systems taking into account different bandwidths within a site and between sites. We consider a new recovery scheme called staged rebuild and compare it with both a direct scheme and a scheme known as intelligent rebuild. To assess the reliability gains achieved by these schemes, we develop an analytical model that incorporates various relevant aspects of storage systems, such as bandwidths, latent sector errors, and failure distributions. The model applies in the context of OpenStack Swift, a widely deployed cloud storage system. Under certain practical system configurations, we establish that order of magnitude improvements in mean time to data loss (MTTDL) can be achieved using these schemes.

show abstract

Section: System Reliabilitymentioning

confidence: 92%