Multi-Queue Fair Queuing
Offered By: USENIX via YouTube
Course Description
Overview
Explore a conference talk on Multi-Queue Fair Queuing (MQFQ), a novel fair and work-conserving scheduler designed for multi-queue systems. Delve into the challenges posed by modern high-speed devices and their multi-queue interfaces, which enable scalable throughput but introduce fairness issues. Learn how MQFQ reformulates classical fair queueing algorithms to accommodate multi-queue designs and implements a scalable solution that balances fairness and synchronization overhead. Discover the impressive performance improvements achieved by MQFQ, including throughput of up to 3.1 Million IOP/s on a single machine and significant reductions in slowdown caused by antagonistic workloads. Gain insights into the technical aspects of MQFQ's implementation in Linux 4.15 and its evaluation using NVMe over RDMA fabric devices.
Syllabus
Intro
Conventional 1/0 Design
Multi-Queue I/O Design
Multi-Queue Fair Queueing
Scalable Implementation
Example: Parallelism Utilization
Evaluation
Conclusion
Taught by
USENIX
Related Courses
Amazon DynamoDB - A Scalable, Predictably Performant, and Fully Managed NoSQL Database ServiceUSENIX via YouTube Faasm - Lightweight Isolation for Efficient Stateful Serverless Computing
USENIX via YouTube AC-Key - Adaptive Caching for LSM-based Key-Value Stores
USENIX via YouTube The Future of the Past - Challenges in Archival Storage
USENIX via YouTube A Decentralized Blockchain with High Throughput and Fast Confirmation
USENIX via YouTube