Within Alluxio, the master processes keep track of global metadata for the file system. This includes file system metadata, block cache metadata, and worker metadata. When a client interacts with the filesystem it must first query or update the metadata on the master processes. Given their central role in the system, master processes can be backed by a highly available, fault tolerant replicated journal. This talk will introduce and compare the two available implementations of this journal in Alluxio, the first using Zookeeper and the more recent version using Raft.
ALLUXIO DAY X 2022
March 3, 2022
Within Alluxio, the master processes keep track of global metadata for the file system. This includes file system metadata, block cache metadata, and worker metadata. When a client interacts with the filesystem it must first query or update the metadata on the master processes. Given their central role in the system, master processes can be backed by a highly available, fault tolerant replicated journal. This talk will introduce and compare the two available implementations of this journal in Alluxio, the first using Zookeeper and the more recent version using Raft.
Video:
Presentation Slides:
Complete the form below to access the full overview:
.png)
Videos
TorchTitan is a proof-of-concept for Large-scale LLM training using native PyTorch. It is a repo that showcases PyTorch's latest distributed training features in a clean, minimal codebase.
In this talk, Tianyu will share TorchTitan’s design and optimizations for the Llama 3.1 family of LLMs, spanning 8 billion to 405 billion parameters, and showcase its performance, composability, and scalability.
As large-scale machine learning becomes increasingly GPU-centric, modern high-performance hardware like NVMe storage and RDMA networks (InfiniBand or specialized NICs) are becoming more widespread. To fully leverage these resources, it’s crucial to build a balanced architecture that avoids GPU underutilization. In this talk, we will explore various strategies to address this challenge by effectively utilizing these advanced hardware components. Specifically, we will present experimental results from building a Kubernetes-native distributed caching layer, utilizing NVMe storage and high-speed RDMA networks to optimize data access for PyTorch training.