Explore strategies for building highly available stateful distributed systems in cloud environments in this 45-minute conference talk from NDC Oslo 2020. Delve into the complexities of state management, comparing stateful and stateless applications, and learn effective approaches for ensuring high availability in Azure and AWS. Examine scaling techniques, multi-region deployments, and the importance of machine placement. Discover tools and services for managing state, including a live demonstration of Azure SQL failover and a comparison with Cosmos DB. Gain insights into storage solutions, queues, and their high-availability features. Understand Service Level Agreements (SLAs) and learn to calculate them for your systems. Leave equipped with practical knowledge to build reliable architectures and make your stateful distributed systems highly available in cloud environments.
Overview
Syllabus
Intro
Panther Science Fiction PHILIP K. DICK
STRATEGY
State - computer science
State, state everywhere
Fully Stateless
Stateful or Stateless?
Fully Stateful
High availability - strategi
Scaling
Multi-region - basic units
Multi-region Azure vs AWS
VM Placement Strategies
Why should we care?
So let's scale Zookeepers!
Highly Available Cloud Services
Database
Scalability
Active Geo-Replication
Failover groups
Demo time!
Azure SQL - my experience
But there is a hope...
Amazon RDS
Cosmos DB
Amazon DynamoDB
Locally-redundant storage
Read-access GRS
Azure Storage V2
Azure queue
Storage - experience
Proxy
Dynamic approach
SLA - Service Level Agreement
Uptime and downtime with 99.9 % SLA
Uptime and downtime with 99.999 % SLA
Uptime approach
So let's go multiregion!
Multiregion SLA
Call to action!
Slides and code
Taught by
NDC Conferences
