On-Demand Videos
Coupang is a leading e-commerce company in South Korea, with over 50,000 employees and $20+ billion in annual revenue. Coupang's AI platform team builds and manages a large-scale AI platform in AWS for machine learning engineers to train models that enhance and customize product search results and product recommendations for its 100+ million customers.
As the search and recommendation models evolve, optimizing the underlying infrastructure for AI/ML workloads is essential for the e-commerce business. Coupang's platform team actively sought to improve their model training pipeline to boost machine learning engineers' productivity, publish models to production faster, and reduce operational costs.
Coupang focused on addressing several key areas:
- Shortening data preparation and model training time
- Improving GPU utilization in training clusters in different regions
- Reducing S3 API and egress costs incurred from copying large training datasets across regions
- Simplifying the operational complexity of storage system management
In this tech talk, Hyun Jung Baek, Staff Backend Engineer at Coupang, will share best practices for leveraging distributed caching to power search and recommendation model training infrastructure.
Hyun will discuss:
- How Coupang builds a world-class large-scale AI platform for machine learning engineers to deliver better search and recommendation models
- How adding distributed caching to their multi-region AI infrastructure improves GPU utilization, accelerates end-to-end training time, and significantly reduces cross-region data transfer costs.
- How to simplify platform operations and to easily deploy the same architecture to new GPU clusters.
About the Speaker
Hyun Jung Baek is a Staff Backend Engineer at Coupang.
Deepseek’s recent announcement of the Fire-flyer File System (3FS) has sparked excitement across the AI infra community, promising a breakthrough in how machine learning models access and process data.
In this webinar, an expert in distributed systems and AI infrastructure will take you inside Deepseek 3FS, the purpose-built file system for handling large files and high-bandwidth workloads. We’ll break down how 3FS optimizes data access and speeds up AI workloads as well as the design tradeoffs made to maximize throughput for AI workloads.
This webinar you’ll learn about how 3FS works under the hood, including:
✅ The system architecture
✅ Core software components
✅ Read/write flows
✅ Data distribution/placement algorithms
✅ Cluster/node management and disaster recovery
Whether you’re an AI researcher, ML engineer, or infrastructure architect, this deep dive will give you the technical insights you need to determine if 3FS is the right solution for you.
.png)
Dataproc is Google’s managed Hadoop and Spark platform. In this talk, we will showcase how to swiftly build a hybrid cloud data platform with Alluxio and Presto and migrate data seamlessly.
Today, many people run deep learning applications with training data from separate storage such as object storage or remote data centers. This presentation will demo the Intel Analytics Zoo + Alluxio stack, an architecture that enables high performance while keeping cost and resource efficiency balanced without network being I/O bottlenecked.
Intel Analytics Zoo is a unified data analytics and AI platform open-sourced by Intel. It seamlessly unites TensorFlow, Keras, PyTorch, Spark, Flink, and Ray programs into an integrated pipeline, which can transparently scale from a laptop to large clusters to process production big data. Alluxio, as an open-source data orchestration layer, accelerates data loading and processing in Analytics Zoo deep learning applications.
This talk, we will go over:
- What is Analytics Zoo and how it works
- How to run Analytics Zoo with Alluxio in deep learning applications
- Initial performance benchmark results using the Analytics Zoo + Alluxio stack
Video: Presentation Slides: Deep Learning in the Cloud at Scale: A Data Orchestration Story from Alluxio, Inc.
Nowadays, cloud native environments have attracted lots of data-intensive applications deployed and ran on them, due to the efficient-to-deploy and easy-to-maintain advantages provided by cloud native platforms and frameworks such as Docker, Kubernetes. However, cloud native frameworks does not provide the data abstraction support to the applications natively. Therefore, we build Fluid project, which co-orchestrate data and containers together. We use Alluxio as the cache runtime inside Fluid to warm up hot data. In this report, we will introduce the design and effects of the Fluid project.
Unisound focuses on Artificial Intelligence services for the Internet of Things. It is an artificial intelligence company with completely independent intellectual property rights and the world’s top intelligent voice technology. Atlas is the Deep Learning platform within Unisound AI Labs, which provides deep learning pipeline support for hundreds of algorithm scientists. This talk shares three real business training scenarios that leverage Alluxio’s distributed caching capabilities and Fluid’s cloud native capabilities, and achieve significant training acceleration and solve platform IO bottlenecks. We hope that the practice of Alluxio & Fluid on Atlas platform will bring benefits to more companies and engineers.
Data and Machine Learning (ML) technologies are now widespread and adopted by literally all industries. Although recent advancements in the field have reached an unthinkable level of maturity, many organizations still struggle with turning these advances into tangible profits. Unfortunately, many ML projects get stuck in a proof-of-concept stage without ever reaching customers and generating revenue. In order to effectively adopt ML technologies, enterprises need to build the right business cases as well as to be ready to face the inevitable technical challenges. In this talk, we will share some common pitfalls, lessons learned, and engineering practices, faced while building customer-facing enterprise ML products. In particular, we will focus on the engineering that delivers real-time audience insights everyday to thousands of marketers via the Helixa’s market research platform.
During the talk you will learn:
- An overview of the Helixa ML end-to-end system
- Useful engineering practices and recommended tools (PyData stack, AWS, Alluxio, scikit-learn, tensorflow, mlflow, jupyter, github, docker, Spark, to name a few..)
- The R&D workflow and how it integrates with the production system
- Infrastructure considerations for scalable and cheap deployment, monitoring, and alerting
- How to leverage modern cloud serverless architectures for data and machine learning applications
Enterprises everywhere are racing to build the optimal analytics stack for creating repeatable success with predictive analytics, machine learning, and data applications. Cloud data platforms like data warehouses and data lakes are foundational elements of these software stacks and their associated data pipelines. But existing SQL query methods against these data platforms have repeatedly demonstrated disappointing performance and scaling due to poor concurrency.
In this presentation, we will discuss the use of the intelligent precomputation capabilities of Kyligence Cloud as a means of delivering on the promise of pervasive analytics at scale with massive concurrency and sub-second query latencies on large datasets in the cloud.
Kyligence, with our partner Alluxio, sits between the data platform and the processing layer. Kyligence Cloud delivers precomputed datasets for OLAP queries, BI dashboards, and machine learning applications.
In most of the distributed storage systems, the data nodes are decoupled from compute nodes. This is motivated by an improved cost efficiency, storage utilization and a mutually independent scalability of computation and storage. While this consideration is indisputable, several situations exist where moving computation close to the data brings important benefits. Whenever the stored data is to be processed for analytics purposes, all the data needs to be repeatedly moved from the storage to the compute cluster, which leads to reduced performance.
In this talk, we will present how using Alluxio computation and storage ecosystems can better interact benefiting of the “bringing the data close to the code” approach. Moving away from the complete disaggregation of computation and storage, data locality can enhance the computation performance. During this talk, we will present our observations and testing results that will show important enhancements in accelerating Spark Data Analytics on Ceph Objects Storage using Alluxio.
At PayPal & any other data driven enterprise – data users & applications work with a variety of data sources (RDBMS, NoSQL, Messaging, Documents, Big Data, Time Series Databases), compute engines (Spark, Flink, Beam, Hive), languages (Scala, Python, SQL) and execution models (stream, batch, interactive) to process petabytes of data. Due to this complex matrix of technologies and thousands of datasets, engineers spend considerable time learning about different data sources, formats, programming models, APIs, optimizations, etc. which impacts time-to-market (TTM).
To solve this problem and to make product development more effective, PayPal Data Platforms developed “Gimel”, an open source, unified analytics data platform which provides access to any storage through a single unified data API and SQL, which are powered by a centralized data catalog.
In this talk, Baolong Mao from Tencent will share his experience in developing Apache Ozone under file system, showing how to create a new Under File System in a few steps with minimal lines of code.
JD.com is one of the largest e-commerce corporations. In big data platform of JD.com, there are tens of thousands of nodes and tens of petabytes off-line data which require millions of spark and MapReduce jobs to process everyday. As the main query engine, thousands of machines work as Presto nodes and Presto plays an import role in the field of In-place analysis and BI tools. Meanwhile, Alluxio is deployed to improve the performance of Presto. The practice of Presto & Alluxio in JD.com benefits a lot of engineers and analysts.
Data platforms span multiple clusters, regions and clouds to meet the business needs for agility, cost effectiveness, and efficiency. Organizations building data platforms for structured and unstructured data have standardized on separation of storage and compute to remain flexible while avoiding vendor lock-in. Data orchestration has emerged as the foundation of such a data platform for multiple use cases all the way from data ingestion to transformations to analytics and AI.
In this keynote from Haoyuan Li, founder and CEO of Alluxio, we will showcase how organizations have built data platforms based on data orchestration. The need to simplify data management and acceleration across different business personas has given rise to data orchestration as a requisite piece of the modern data platform. In addition, we will outline typical journeys for realizing a hybrid and multi-cloud strategy.