AI Powers Databases and Databases Power AI

Attend this event to learn AI and Databases to the power of 2: Professor Andy Pavlo will explain how OtterTune uses Machine Learning to automatically tune your database. You will also learn how the distributed database - TiDB and TiKV empowers the implementation of SHAREit’s AI Platform & Recommendation System.

Speaker:
Andy Pavlo, Associate Professor at Carnegie Mellon University, CEO of OtterTune.
Andy is an Associate Professor of Databaseology in the Computer Science Department at Carnegie Mellon University. He is also the co-founder of OtterTune (https://ottertune.com).

Marshall Zhu, big data director, SHAREit
Marshall is the big data director of SHAREit. SHAREit Group is a global mobile internet application development and digital services company. The applications suite of SHAREit has been installed by nearly 2.4 billion users worldwide. The business network has reached over 200 countries and regions in 45 different languages.

Host:
Jonathan Baker, Head of Developer Relations and Community for North America at PingCAP
Jonathan is the Head of Developer Relations and Community for North America at PingCAP. He has spent over 20 years working in Developer Relations, managing and directly influencing and communicating with developers.

Agenda:
How to Use Machine Learning to Stop Overpaying Jeff Bezos for Your Databases - Andy Pavlo
AI Platform & Recommendation System Implementation in SHAREit - Marshall Zhu

Detailed Description:
TITLE:How to Use Machine Learning to Stop Overpaying Jeff Bezos for Your Databases - Andy Pavlo
ABSTRACT:
Database management systems (DBMS) expose dozens of configurable knobs that control their runtime behavior. Setting these knobs correctly for an application's workload can improve the performance and efficiency of systems like PostgreSQL and MySQL. But such tuning requires considerable efforts from experienced administrators. Because it is so difficult to tune these systems manually, too many people end up overpaying cloud vendors to run their databases. This problem has led to research on using machine learning (ML) to devise strategies to automatically optimize DBMS knobs for any application.

In this talk, Professor Andy Pavlo will present an overview of the OtterTune database tuning service. OtterTune uses ML to generate and install optimized configurations by observing DBMS's workload and training recommendation models that select better knob values. I will also discuss the challenges one must overcome to deploy an ML-based service for DBMSs and highlight the insights we learned from real-world installations of OtterTune.

TITLE: AI Platform & Recommendation System Implementation in SHAREit - Marshall Zhu
ABSTRACT:
Nowdays, Machine Learning has become a toolkit that is easy to reach out for every start-up company. But what makes AI powerful is not a typical model, it relies on efficient iterations in the AI workflow, including preparation, build, and deployment. Meanwhile, the recommendation system acts as one of the most important AI scenarios, it's inevitable to deal with many handshakes with the platform.

In this talk, Marshall will share our comprehension of AI workflow in SHAREit. Then he will present what hurts us most in the workflow, what challenges we face, what kinds of capabilities we have and how we build the platform from scratch. Marshall will also sit on the user side to describe how our recommendation system communicates with the AI Platform. Furthermore, he will describe how TiKV and TiDB help us to achieve these goals.