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

Hybrid storage systems (HSS) use multiple different storage devices toprovide high and scalable storage capacity at high performance. Recent researchproposes various techniques that aim to accurately identifyperformance-critical data to place it in a "best-fit" storage device.Unfortunately, most of these techniques are rigid, which (1) limits theiradaptivity to perform well for a wide range of workloads and storage deviceconfigurations, and (2) makes it difficult for designers to extend thesetechniques to different storage system configurations (e.g., with a differentnumber or different types of storage devices) than the configuration they aredesigned for. We introduce Sibyl, the first technique that uses reinforcementlearning for data placement in hybrid storage systems. Sibyl observes differentfeatures of the running workload as well as the storage devices to makesystem-aware data placement decisions. For every decision it makes, Sibylreceives a reward from the system that it uses to evaluate the long-termperformance impact of its decision and continuously optimizes its dataplacement policy online. We implement Sibyl on real systems with various HSSconfigurations. Our results show that Sibyl provides 21.6%/19.9% performanceimprovement in a performance-oriented/cost-oriented HSS configuration comparedto the best previous data placement technique. Our evaluation using an HSSconfiguration with three different storage devices shows that Sibyl outperformsthe state-of-the-art data placement policy by 23.9%-48.2%, while significantlyreducing the system architect's burden in designing a data placement mechanismthat can simultaneously incorporate three storage devices. We show that Sibylachieves 80% of the performance of an oracle policy that has complete knowledgeof future access patterns while incurring a very modest storage overhead ofonly 124.4 KiB.

Sibyl: Adaptive and Extensible Data Placement in Hybrid Storage Systems Using Online Reinforcement Learning