Explore the development of a new FDB memory storage engine based on the radix tree data structure in this 17-minute conference talk by Mengran Wang from VMware. Learn about the advantages of using a radix tree for space-efficient key-value storage, particularly in scenarios with significant prefix overlap in keys. Discover how this trie-based KV store compares to the default balanced binary search tree engine, and understand its benefits for time series workloads. Delve into the optimizations implemented, including prefix compression and adaptive registry. Examine performance comparisons using real-life Wavefront workloads, and gain insights into memory usage, key formatting, and node structure. Analyze test results showcasing total memory usage and performance metrics across various benchmarks, including a Wikipedia title test.
A New Radix-Tree Based Memory Storage Engine
Overview
Syllabus
Introduction
Agenda
Key Container
Motivation
Prefix Compression
Demo
Advantages
Optimization
Test Environment
Test Cluster Configuration
Memory Usage Gap
Key Format
Registry
Bottom Line
Node Structure
Optimizations
Data part
Inline data
Adaptive registry
Internodes
Vector
Value
Key Smile
Summary
Note overhead
Performance results
Total Memory Usage
Histogram
Wikipedia Title Benchmark
Thank you
Taught by
Linux Foundation
