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

Recently using machine learning (ML) based techniques to optimize moderndatabase management systems has attracted intensive interest from both industryand academia. With an objective to tune a specific component of a DBMS (e.g.,index selection, knobs tuning), the ML-based tuning agents have shown to beable to find better configurations than experienced database administrators.However, one critical yet challenging question remains unexplored -- how tomake those ML-based tuning agents work collaboratively. Existing methods do notconsider the dependencies among the multiple agents, and the model used by eachagent only studies the effect of changing the configurations in a singlecomponent. To tune different components for DBMS, a coordinating mechanism isneeded to make the multiple agents cognizant of each other. Also, we need todecide how to allocate the limited tuning budget among the agents to maximizethe performance. Such a decision is difficult to make since the distribution ofthe reward for each agent is unknown and non-stationary. In this paper, westudy the above question and present a unified coordinating framework toefficiently utilize existing ML-based agents. First, we propose a messagepropagation protocol that specifies the collaboration behaviors for agents andencapsulates the global tuning messages in each agent's model. Second, wecombine Thompson Sampling, a well-studied reinforcement learning algorithm witha memory buffer so that our framework can allocate budget judiciously in anon-stationary environment. Our framework defines the interfaces adapted to abroad class of ML-based tuning agents, yet simple enough for integration withexisting implementations and future extensions. We show that it can effectivelyutilize different ML-based agents and find better configurations with 1.4~14.1Xspeedups on the workload execution time compared with baselines.

A Unified and Efficient Coordinating Framework for Autonomous DBMS Tuning