English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Dynamic random access memory failures are a threat to the reliability of datacentres as they lead to data loss and system crashes. Timely predictions ofmemory failures allow for taking preventive measures such as server migrationand memory replacement. Thereby, memory failure prediction prevents failuresfrom externalizing, and it is a vital task to improve system reliability. Inthis paper, we revisited the problem of memory failure prediction. We analyzedthe correctable errors (CEs) from hardware logs as indicators for a degradedmemory state. As memories do not always work with full occupancy, access tofaulty memory parts is time distributed. Following this intuition, we observedthat important properties for memory failure prediction are distributed throughlong time intervals. In contrast, related studies, to fit practicalconstraints, frequently only analyze the CEs from the last fixed-size timeinterval while ignoring the predating information. Motivated by the observeddiscrepancy, we study the impact of including the overall (long-range) CEevolution and propose novel features that are calculated incrementally topreserve long-range properties. By coupling the extracted features with machinelearning methods, we learn a predictive model to anticipate upcoming failuresthree hours in advance while improving the average relative precision andrecall for 21% and 19% accordingly. We evaluated our methodology on real-worldmemory failures from the server fleet of a large cloud provider, justifying itsvalidity and practicality.

First CE Matters: On the Importance of Long Term Properties on Memory Failure Prediction