ALL THINGS AI
We’ve compiled a collection of 2023’s most popular content according to our readers. In case you missed anything, here’s your chance to catch up on best practices ebooks, technical blogs, hands on videos, webinars and more!
Mengyu Hu and Chengkun Jia, both from Zhihu’s data platform team, discuss their evolution from HDFS to Alluxio as a high-performance data access layer for LLM training and serving. Alluxio has accelerated model training by 2~3x, increased GPU utilization to 90%, and enabled model deployment every minute instead of hours or days.
Get a comprehensive understanding of data access patterns in a modern AI/ML platform. This white paper discusses the characteristics of data access in each stage of the machine learning pipeline and the solutions that can be used in architecting your data and AI platform.
When training models on ultra-large datasets, one of the biggest challenges is low GPU utilization. These powerful processors are often underutilized due to inefficient I/O and data access. This mismatch between computation and storage leads to wasted GPU resources, low performance, and high cloud storage costs. The rise of generative AI and GPU scarcity is only making this problem worse.
In this webinar, Tarik and Beinan discuss strategies for transforming idle GPUs into optimal powerhouses. They will focus on cost-effective management of ultra-large datasets for AI and analytics.
Watch the Alluxio Enterprise AI end-to-end ML pipeline demo, and see for yourself the significant performance improvements as well as increased GPU utilization! Alluxio’s Solution Engineer Tarik Bennett walks through a short end-to-end machine learning pipeline with Alluxio provisioned or mounted as a local folder for PyTorch dataloader.
Alluxio’s Senior Solutions Engineer Roland Theron shares how Alluxio benefits model training workflows by reducing data loading times, allowing for better utilization of your compute resources.
Discover the easily consumed tuning tips that deliver optimal training speeds at lower costs. Learn how to tune PyTorch performance to achieve lower latency and higher GPU utilization through data loading, data operations, GPU processing, and CPU processing, with lines of code.
Explore the transformative capabilities of the Data Access Layer and how it can simplify and accelerate your analytics and AI workloads. in this new research paper, Kevin Petrie, VP of Research at Eckerson Group, shares the architecture and use cases for a Data Access Layer and how it can help achieve analytics and AI goals with successful performance.
Cost Savings and Optimization
Find out details of our joint project with Uber aimed at optimizing the performance of HDFS DataNodes. The project utilized the Alluxio SDK cache to manage an SSD storage on each DataNode, resulting in improved performance and a better return on investment. Despite the SSD cache occupying only 0.6% of the total disk space, it impressively handles 60% of the overall client traffic.
Build your data platform with reduced cloud egress costs and never be surprised by a bill again. Minimize your data replication with optimized data pipelines and data flow for your architecture.
Learn how Shopee, the leading e-commerce platform in Asia, has successfully leveraged Alluxio to improve Trino query performance by ~55%. In addition, Alluxio enhances developer experience by providing flexible data access through Data APIs.
Unlock the full potential of Trino and transform your data analytics game. Identify bottlenecks and maximize your Trino query performance with configuration settings and session properties.
If you’re using Presto (PrestoDB), check out The Presto Optimization Handbook here.
Got a tech question for the Alluxio Community? Chat with us on Slack!
Be our stargazers on GitHub ⭐
If you like our product, please give it a star on GitHub, and share the goodness!