Distributed Systems

Links

• In Search of Five 9s - Calculating Availability of Complex Systems

• When AWS autoscale doesn't

• It's Time to Move on from Two Phase Commit

• Managing Machines at Spotify

• Notes on Distributed Systems for Young Bloods

• Delivering billions of events per day

• Consistency Models

• UTC is Enough for Everyone, Right?

• Testing Distributed Systems | Curated list of resources on testing distributed systems

• Designing robust and predictable APIs with idempotency

• Distributed architecture concepts I learned while building a large payments system

• The Reactive Manifesto

• The Log: What every software engineer should know about real-time data's unifying abstraction

• Testing microservices

• Event-driven Data or: How I Learned to Stop Living in the Present and Travel Through Time

• Microservices are hard — an invaluable guide to microservices

• Hidden costs of serverless

• Principles of chaos engineering

• Immutable Server

• 10 Lessons from 10 Years of AWS (part 2)

• Serverless computing: economic and architectural impact

• Asynchronus computability theorem

• Databases and Distributed Deadlocks: A FAQ

• There is No Now

• Formal verification of distributed systems with Verdi

• Distributed systems for fun and profit

The cloud native maturity model

Consistency Models

(Summary of Strong consistency models)

Concurrent Histories

With several processes, I write to one common storage, that can be spread across multiple nodes.

Light Cones

There is a time slot for the task to write to acknowledge the write. Same for read access.

Linearizability

If a write task was acknowledged, the result is visible for all reading clients.

Sequential Consistency

If something was given to the system in a certain sequence, the visibility will be in the same sequence. Example: posts (A, then B) to a social network are not visible immediately. But if A becomes visible, only after that, B will become visible.

Causal Consistency

Only if operations are dependent from another, they will be given in the correct sequence. For example can read access only be allowed after certain conditions are met.

Serializable Consistency

Is weak because reads and writes can travel to the past and to the future. Is strong because it requires a linearization and certain conditions. The reading history is determined.