Machine-Learned Indexes - Research from Google | Software Architecture Conference

Abstract

Modern data processing systems are designed to be general purpose, in that they can handle a wide variety of different schemas, data types, and data distributions, and aim to provide efficient access and computation over this data. This “one-size-fits-all” nature results in systems that do not take advantage of the unique characteristics of each application, data of the user, or workload. However, ignored in these old systems’ design: machine learning excels at understanding and adapting to particular datasets. We present here a vision (with evidence) for the future of data processing systems: through learning models of the application, data, and workload, we can redesign and customize nearly every component of data processing systems. We will do a deep-dive into understanding how traditional index structures can be reframed as machine learning problems, and that by doing so, and through careful model design and code synthesis, we are able to outperform cache-optimized B-Trees by up to 70% in speed while saving an order-of-magnitude in memory over several real-world data sets. Building on these same modeling techniques, we find that we can achieve improvements in sorting, multi-dimensional indexing, and query optimization, all areas that have historically been the domain of traditional discrete algorithms and complex systems engineering.

Speaker: Alex Beutel

Senior Research Scientist @Google

Alex Beutel is a Senior Research Scientist in the Google Brain SIR team working on neural recommendation, fairness in machine learning, and ML for Systems. He received his Ph.D. in 2016 from Carnegie Mellon University’s Computer Science Department, and previously received his B.S. from Duke University in computer science and physics. His Ph.D. thesis on large-scale user behavior modeling, covering recommender systems, fraud detection, and scalable machine learning, was given the SIGKDD 2017 Doctoral Dissertation Award Runner-Up. He received the Best Paper Award at KDD 2016 and ACM GIS 2010, was a finalist for best paper in KDD 2014 and ASONAM 2012, and was awarded the Facebook Fellowship in 2013 and the NSF Graduate Research Fellowship in 2011. More details can be found at alexbeutel.com.

Find Alex Beutel at

Speaker page

Speaker: Tim Kraska

Associate Professor @MIT & Visiting Researcher @Google

Find Tim Kraska at

Speaker page

Tracks

Monday, 24 June

Architectures You've Always Wondered About

Hard earned lessons from names you know on scalability, reliability, throughput, and performance.
Modern CS in the Real World

Thoughts pushing software forward, including consensus, CRDTs, formal methods, & probabilistic programming.
Machine Learning for Developers

Machine learning is more approachable than ever. Learn techniques and use cases with PyTorch, Keras and TensorFlow that will become foundational for moderan application developers.
Developing/Optimizing Clients for Developers

Electron, Flutter, JavaScript, GraphQL, Node/NPM. Web Workers, Wasm, WebVR, Speech/Alexia/Pipeline. More topics than room in this modern UI tech track.
Human Systems: Hacking the Org

Innovative approaches to organizational leadership and management to build unique company cultures.

Tuesday, 25 June

Microservices / Serverless (Patterns & Practices)

Evolving, observing, persisting, operating, and building modern architectures.
Modern Java Innovations

6 month cadence, cloud-native deployments, scale, Graal, Kotlin, and beyond. Learn how the role of Java as it is evolves for the next 20 years.
Data Engineering for the Bold

Explore the power of data systems & architectures
Non-Technical Skills for Technical Folks

Good engineers are great programmers. Amazing engineers are great collaborators.
Trust, Safety, & Security

Privacy, confidentiality, safety, security: Learnings from the frontlines

Wednesday, 26 June

Architecting for Success when Failure is Guaranteed

More than just building software, building deployable production ready software in the face of guaranteed failure.
21st Century Languages

Lessons learned from building languages like Rust, Go-lang, Swift, Kotlin, and more.
Building High-Performing Teams

What “high-performing team” means and how to build one effectively depends on context. This track will share different experiences of building high-performing teams in order to highlight how different contexts lead to different solutions but also what typically stays the same because we’re still dealing with humans trying to work together. How do different forces affect the building of high-performing teams.
Software Defined Infrastructure: Kubernetes, Service Meshes, & Beyond

Deploying, scaling, managing your services is undifferentiated heavy lifting. Hear stories, learn techniques, and dive deep into software infrastructure.
High-Performance Computing: Lessons from FinTech & AdTech

Killing latency and getting the most out of your hardware.

Schedule

Presentation: Machine-Learned Indexes - Research from Google

Track: Architectures You've Always Wondered About

Location: Broadway Ballroom North Center, 6th fl.

Duration: 11:50am - 12:40pm

Day of week: Monday

More talks on:

Share this on: