How We Saved 5x Migrating from Aurora to FoundationDB

Discover how Segment achieved a 5x cost reduction by migrating from AWS Aurora to FoundationDB in this informative conference talk. Explore the journey of Segment's identity resolution system, a critical component in translating customer interactions into Personas models. Learn about the architecture behind Segment's infrastructure, the decision-making process for selecting FoundationDB, and the significant benefits realized through this migration. Gain insights into the specific operational challenges faced during the transition and the strategies employed to overcome them. Delve into topics such as CDI, identity graphs, data models, workload profiles, cluster evolution, and infrastructure setup. Understand the importance of exactly once delivery and the pros and cons of deduplication in this context. This presentation offers valuable lessons for organizations considering similar database migrations to improve performance and reduce costs.

Intro
What is CDI?
Segment Timeline
Identity Resolution Use Cases
Problem
Identity Graph
Mappings
General Data Model
"Object" KV Kind
Mapping Object
Segment ID Index
Merge Object
Workload Profile Needs
Personas Data Flow
Aurora works, but...
Some Baseline Aurora Costs
Example Realistic Workload
Tuning Cookbook
Cluster Evolution
Infrastructure Overview
Terracode-foundationdb / cluster
AMI Launch Config / User Data
data.mount
configure-fdb
fdb-discovery
fdb-backup
Gameday & Chaos
Exactly Once Delivery
Dedupe Pros / Cons