MegaScale - Scaling Large Language Model Training to More Than 10,000 GPUs
Innovative system for training large language models on 10,000+ GPUs, addressing efficiency and stability challenges through full-stack optimization and advanced diagnostic tools.
LoLKV - Logless, Linearizable, RDMA-based Key-Value Storage System
Innovative key-value storage system using lock-free design for concurrent updates, achieving higher throughput and lower latency than state-of-the-art alternatives.
The Bedrock of Byzantine Fault Tolerance: A Unified Platform for BFT Protocols - NSDI '24
Unified platform for analyzing, implementing, and experimenting with Byzantine Fault-Tolerant protocols, exploring design dimensions and trade-offs to navigate the complex BFT landscape.
Resiliency at Scale: Managing Google's TPUv4 Machine Learning Supercomputer
Explore Google's TPUv4 supercomputer infrastructure, focusing on fault resiliency, dynamic reconfiguration, and high system availability for machine learning training at massive scale.
Modularity in PVS
Explore modularity concepts in PVS theorem proving, enhancing code organization and reusability for complex formal verification projects.
Optimizing a Ray Tracer by Building a Bounding Volume Hierarchy (BVH)
Optimize ray tracing performance by building a Bounding Volume Hierarchy (BVH). Learn techniques like GPU-friendly data structures, iterative traversal, and surface area heuristics to render complex scenes efficiently.
Event-Driven Autoscaling for Serverless Java on Kubernetes
Explore event-driven autoscaling for serverless Java using KEDA and Knative on Kubernetes, focusing on Apache Kafka metrics for efficient scaling beyond traditional resource-based methods.
System Introductions I - HOL - Lecture 1
Introduction to the HOL theorem prover, exploring its features and applications in formal mathematics and computer science verification.
Even More on HOL Light - Lecture 3
Explore advanced topics in HOL Light theorem proving, building on previous sessions to deepen understanding and practical application of this powerful system.
More on HOL Light - Lecture 2
Explore advanced concepts and applications of HOL Light theorem prover with expert Freek Wiedijk in this in-depth continuation of the series.
Lean-HoG: Incorporating a Database of Graphs into a Proof Assistant
Explore the integration of a graph database into a proof assistant, enhancing mathematical reasoning capabilities and streamlining complex proofs in formal verification.
Complexity in Event-Driven Architecture - Managing Challenges and Realizing Potential
Explore how event-driven architecture enhances agility and innovation while managing complexity. Learn strategies for evolutionary architectures, governance, and harnessing EDA's full potential in dynamic business environments.
Architecture Modernization: Aligning Software, Strategy, and Structure
Holistic approach to modernizing legacy architectures by aligning business strategy, product vision, domain boundaries, team organization, and software architecture for innovation and competitive advantage.
How Prometheus Halved Its Memory Usage
Discover how Prometheus significantly reduced its memory usage, enabling efficient monitoring of millions of metrics. Learn about data structures, profiling, and impacts on related projects.
Real-Time Argo Rollouts Analytics Powered by Notification Engine
Discover how Intuit leverages notification engine for real-time Argo Rollouts analytics, enabling efficient management of applications across clusters and rapid identification of issues at scale.
Scaling Argo Events for Enterprise Scheduling - A Case Study from Intuit
Explore Intuit's journey in scaling Argo Events for enterprise scheduling, including challenges, solutions, and performance improvements for enhanced scalability and reliability.
Scaling to Thousands of Data and CI/CD Pipelines Using Argo and Virtual Clusters
Explore how Pipekit and Loft Labs scale Kubernetes clusters for Argo Workflows, addressing challenges in tenant isolation, automated provisioning, and observability for data workloads at scale.
Cross Cluster Execution of Argo Workflows
Explore cross-cluster Argo Workflows execution, discussing benefits, challenges, and potential solutions for enhanced manageability and resiliency in Kubernetes environments.
Training and Evaluating LLaMA2 Models with Argo Workflows and Hera
Comprehensive guide to train and evaluate LLaMA2 models using Argo Workflows and Hera SDK. Covers setup, data preprocessing, model training, and performance assessment with cloud-native tools.
How to Develop Like a Senior Software Engineer - Key Principles and Real-World Application
Discover key principles of senior-level software engineering through a real-world repository upgrade, applying solid concepts to elevate any project's quality and efficiency.
Tracing the Journey into Distributed Tracing
Explore the challenges and successful strategies for implementing distributed tracing in large organizations, with insights on onboarding new technologies effectively.