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

Meeting desired application deadlines in cloud processing systems such asMapReduce is crucial as the nature of cloud applications is becomingincreasingly mission-critical and deadline-sensitive. It has been shown thatthe execution times of MapReduce jobs are often adversely impacted by a fewslow tasks, known as stragglers, which result in high latency and deadlineviolations. While a number of strategies have been developed in existing workto mitigate stragglers by launching speculative or clone task attempts, none ofthem provides a quantitative framework that optimizes the speculative executionfor offering guaranteed Service Level Agreements (SLAs) to meet applicationdeadlines. In this paper, we bring several speculative scheduling strategiestogether under a unifying optimization framework, called Chronos, which definesa new metric, Probability of Completion before Deadlines (PoCD), to measure theprobability that MapReduce jobs meet their desired deadlines. We systematicallyanalyze PoCD for popular strategies including Clone, Speculative-Restart, andSpeculative-Resume, and quantify their PoCD in closed-form. The resultilluminates an important tradeoff between PoCD and the cost of speculativeexecution, measured by the total (virtual) machine time required underdifferent strategies. We propose an optimization problem to jointly optimizePoCD and execution cost in different strategies, and develop an algorithmicsolution that is guaranteed to be optimal. Chronos is prototyped on HadoopMapReduce and evaluated against three baseline strategies using bothexperiments and trace-driven simulations, achieving 50% net utility increasewith up to 80% PoCD and 88% cost improvements.

Chronos: A Unifying Optimization Framework for Speculative Execution of Deadline-critical MapReduce Jobs