Relational Databases vs NoSQL
| Relational Databases | NoSQL |
| Use tables/rows/columns | Don't Use tables/rows/columns |
| Need a predefind schema / complicated to change structure | Schema less / easy change structure |
| Slow queries when joining multiple tables | Fast queries |
| Vertically scalable + More expernsive | Horizontal scalable + Cheaper |
| Guarantee ACID transactions | Most don't support ACID transactions |
Types of NoSQL Databases
- Key-value database
- Document database
- Column family database
- Graph database
Key-value databases
- Such as map data structure
- Is a simplest NoSQL databases
- Set / Get values with associated key
Key must be unique and not good option to have long keys because they will use more memory.
Advantages of key-value databases
Very simple
- Key-value tuple
- Not defined schema
- Basic operations such as
Put,GetandDelete
Very flexible
- Allow changes in data type such as
User12 = 0111, User12 = "Giza" - Add additional attributes such as
User12 = {"lang": "en:US"}, User12 = {"lang": "en:US", "Color": "green"}
- Allow changes in data type such as
- Information stored in memory
- Fast read & write
- Scalability
- Sharding => distributes different parts of the data accross multiple servers
Limitations of key-value databases
- Just search by key
- Problem if we don't the key !!!!!!!!
- Some key-value databases added functionalities to avoid this problem
- Search by value
- Add secondary index
- Search by serveral key simultaneously
- Not complex query
Suitable cases for value-key databases
Shopping carts
- Key => UserID
- Value => shopping cart information
Real time recommendations
- Key => UserID
- Value => items that user like it
The most popular key-value databases is "Redis"