Explore serverless performance optimization techniques in this 27-minute conference talk. Dive into the challenges of current serverless platforms, including cold starts, platform overhead, and inconsistent performance. Learn about various optimization strategies such as function reuse, autoscaling, resource pooling, function locality, and predictive scheduling. Examine how to fine-tune the Fission FaaS platform to achieve specific goals in latency, throughput, resource utilization, and cost. Gain insights into performance and cost models, runtime components, and advanced techniques like function prewarming and workflow optimization. Discover the future opportunities and challenges in making Function-as-a-Service (FaaS) even faster and more efficient.
Overview
Syllabus
Introduction
Performance Model
Cost Model
Runtime Components
Performance
Cold Start
Techniques
First Optimization
Tradeoff
Sharing
Insight
Research pooling
Rebalancing
Research Pool
Prefetching
Hierarchy of Layers
Function Prewarming
Visualizing Function Prewarming
Static Information
fission workflows
prewarm
Summary
Taught by
CNCF [Cloud Native Computing Foundation]
