November 6, 2022 6:00 PM PST

This document summarizes a presentation on NoSQL databases, highlighting their differences from traditional SQL databases, various types of NoSQL databases, and their specific use cases. The discussion covers data modeling, query features, and scaling strategies for different NoSQL systems, including document databases, columnar databases, key-value stores, graph databases, and time series databases.

Presenter: Marco

Key Topics Discussed

Introduction to NoSQL

Traditional SQL databases store data in tables, which can hinder development productivity due to the need for data normalization and complex joins.
NoSQL databases emerged to address these limitations, allowing for more flexible data storage and retrieval.

Types of NoSQL Databases

Document Databases
- Data is stored in JSON or BSON format.
- Supports nested queries.
- MongoDB is the most popular document database.
- Key differences from RDBMS:
  - Collections vs. tables
  - Documents vs. rows
  - _id vs. rowid
  - Reference/embedding vs. joins
  - Consistency through replica sets
  - High availability via leader-follower architecture
  - JSON-like query language
  - Sharding for scaling reads and writes.
Columnar Databases
- Data stored in cells with unlimited columns per row.
- Cassandra is a notable example.
- Key differences from RDBMS:
  - Keyspace vs. database
  - Column family vs. table
  - Consistency through commit logs and memtables
  - Atomic transactions at the single row level.
  - Sharding for scaling.
Key-Value Databases
- Simplest form of NoSQL databases.
- Redis is the most popular key-value store.
- Key differences from RDBMS:
  - Key-value pairs vs. rows
  - Consistency applicable to a single key.
  - Supports transactions with multi/watch/exec and Lua scripts.
Graph Databases
- Composed of nodes and edges, representing relationships.
- Neo4j is a prominent graph database.
- Key features:
  - ACID-compliant and implemented in Java.
  - Supports billions of entities and is schemaless.
  - Leader-follower consistency model.
  - Does not support sharding.
Time Series Databases
- Used for capturing metrics over time.
- InfluxDB is a well-known time series database.
- Key features:
  - Write-ahead log (WAL) and time-structured merge tree.
  - InfluxQL is SQL-like.
  - Supports horizontal scaling but is not a transactional database.

Use Cases for NoSQL Databases

MongoDB: E-commerce applications (e.g., orders, blogging).
Key-Value Stores: Caching user profiles, session information, and shopping cart data.
Graph Databases: Recommendation systems and location-based services.
Cassandra: Event logging.

Additional Concepts

Partitioning and Replication: Strategies for distributing data across nodes to enhance availability and performance.
Indexing: Different databases utilize various indexing strategies to optimize query performance.

Q&A Session

Clarification on the purpose of column families in columnar databases.
Discussion on the amount of data a single database can hold, which depends on the disk size of the hosting computer.

This summary encapsulates the essential points discussed during the presentation on NoSQL databases, providing insights into their architecture, functionalities, and practical applications.