Twitter

Generating unique ID seems to be a simple task, but it is not in a high-load distributed systems!

This topic consists of:

Understanding the requirements and why it is a complicated task
Possible solutions:
- Multi-master replication
- Universally unique identifier (UUID)
- Ticket server
- Twitter SNOWFLAKE approach (seems to be the best one!)
Details:
- Timestamp
- Sequence number
Other issues
- Clock synchronization
- Section length tuning
- High availability

These items are disclosed in a very interesting Chapter 7 of the book:

Earlier this year the book club of our company has studied excellent book:

Martin Kleppmann - Designing Data-Intensive Applications

This is the best book I have read about building complex scalable software systems. 💪

As usually (to better learn) I prepared an overview and mind-map.

Chapter 1:

Building blocks of the apps
What is Reliability, Scalability and Maintainability. Examples and definitions.
- Faults and Failures
- Performance, Load, Latency and Response Time
- Operability, Simplicity, Evolvability
Why you should randomly kill your servers 😅
How Twitter delivers 12,000 tweets per second to 300,000 readers per second. (VERY interesting!)
How much money Amazon loses for each 100ms delay in their response time
How to quickly calculate percentiles for monitoring response time in PROD