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

Datacenters execute large computational jobs, which are composed of smallertasks. A job completes when all its tasks finish, so stragglers -- rare, yetextremely slow tasks -- are a major impediment to datacenter performance.Accurately predicting stragglers would enable proactive intervention, allowingdatacenter operators to mitigate stragglers before they delay a job. While muchprior work applies machine learning to predict computer system performance,these approaches rely on complete labels -- i.e., sufficient examples of allpossible behaviors, including straggling and non-straggling -- or strongassumptions about the underlying latency distributions -- e.g., whetherGaussian or not. Within a running job, however, none of this information isavailable until stragglers have revealed themselves when they have alreadydelayed the job. To predict stragglers accurately and early without labeledpositive examples or assumptions on latency distributions, this paper presentsNURD, a novel Negative-Unlabeled learning approach with Reweighting andDistribution-compensation that only trains on negative and unlabeled streamingdata. The key idea is to train a predictor using finished tasks ofnon-stragglers to predict latency for unlabeled running tasks, and thenreweight each unlabeled task's prediction based on a weighting function of itsfeature space. We evaluate NURD on two production traces from Google andAlibaba, and find that compared to the best baseline approach, NURD produces2--11 percentage point increases in the F1 score in terms of predictionaccuracy, and 2.0--8.8 percentage point improvements in job completion time.

NURD: Negative-Unlabeled Learning for Online Datacenter Straggler Prediction