Heterogeneity-Aware Cluster Scheduling Policies for Deep Learning Workloads
Explores heterogeneity-aware scheduling for deep learning workloads, introducing Gavel to optimize resource allocation in clusters with diverse accelerators, improving efficiency and performance.
UNVEIL - A Large-Scale, Automated Approach to Detecting Ransomware
Automated approach to detect ransomware through dynamic analysis, focusing on file tampering and desktop changes. Improves detection of evasive ransomware unnoticed by traditional antimalware solutions.
Stealing Machine Learning Models via Prediction APIs
Explores model extraction attacks on machine learning systems, demonstrating vulnerabilities in publicly accessible ML interfaces and proposing potential countermeasures to protect confidential models.
Virtual U - Defeating Face Liveness Detection by Building Virtual Models from Your Public Photos
Innovative approach to bypass face authentication using 3D models created from social media photos, exploiting virtual reality to simulate liveness and fool security systems.
Building a Billion User Load Balancer
Discover Facebook's innovative load balancing techniques for handling billions of users, improving performance, managing capacity, and enhancing reliability at a global scale.
MySQL Automation at Facebook Scale
Explore Facebook's MySQL automation systems for managing thousands of servers across data centers, including design, architecture, and real-world challenges faced at massive scale.
BetrFS - A Right-Optimized Write-Optimized File System
Explores BetrFS, an innovative file system using write-optimized indexes, showcasing improved performance for microdata operations and large sequential I/O compared to traditional file systems.
F2FS - A New File System for Flash Storage
Explore F2FS, a Linux file system optimized for flash storage. Learn its design, data structures, and algorithms, and see how it outperforms EXT4 in various scenarios on mobile and server systems.
SRE in the Small and in the Large
Explore SRE principles applicable to organizations of all sizes, from startups to tech giants. Learn key practices for system reliability, scalability, and efficient operations.
Network-Based LUKS Volume Decryption with Tang
Network-based LUKS volume decryption using Tang: a solution for secure, automated server disk encryption at scale. Enables unattended boot in secure environments while protecting data at rest from unauthorized access.
The Devopsification of Windows Server 2016
Explore Windows Server 2016's DevOps-centric features, including NanoServer, DSC, OneGet, and containers. Learn how these innovations enable cloud-paced workflows and enhance security, deployment, and management.
Building a Billion User Load Balancer
Discover Facebook's innovative load balancing techniques for handling over 1.3 billion users, improving performance, managing capacity, and enhancing reliability across global infrastructure.
Scaling Shopify's Multi-Tenant Architecture across Multiple Datacenters
Explore Shopify's journey from a single-database Rails app to a multi-datacenter setup, focusing on scalability, resiliency, and disaster recovery strategies for multi-tenant architectures.
Doorman - Global Distributed Client Side Rate Limiting
Explore Google's Doorman system for global distributed client-side rate limiting, coordinating resource usage across multiple clients to prevent capacity overload.
Food for Thought - What Restaurants Can Teach Us about Reliability
Explore how restaurant operations parallel computer system reliability, drawing insights from dining experiences to enhance understanding of complex system management and fault tolerance.
Cache for Cash - Speeding Up Production with Kafka and MySQL binlog
Optimize high-scale production systems with Kafka and MySQL binlog for fresh, fast data distribution. Learn to balance caching tradeoffs and improve performance across thousands of services.
Taking Control of Metrics Growth and Cardinality - Tips for Maximizing Your Observability
Strategies for managing metrics growth and cardinality in cloud-native environments, focusing on best practices, KPIs, and team collaboration to enhance observability and streamline remediation processes.
Beyond Goldilocks Reliability
Explore reliability engineering beyond traditional SRE practices. Learn about concrete models, underlying mechanisms, and new strategies to enhance service reliability and tackle complex challenges.
Hard Problems We Handle in Incidents but Aren't Recognized
Exploring hidden challenges in incident management, including diagnostic work, coordination costs, and decision-making dilemmas, with insights for improved recognition and handling.
Nine Questions to Build Great Infrastructure Automation Pipelines
Explore nine key questions for building effective infrastructure automation pipelines, focusing on modular design, intent-driven approaches, and seamless tool integration for continuous delivery.
Automating Performance Tuning with Machine Learning
Explore machine learning-driven automation for optimizing Kubernetes microservices and JVM settings, enhancing performance, efficiency, and cost-effectiveness in complex tech stacks.