Relational Databases

• Imposes Tabular style organization:

• Everything is a table.

• Rows are often referred to as Records.

• The data stored in them are well defined.

• NoSQL are hard to query, whereas RDBMS use powerful SQL.

• SQL follows ACID transactions:

  • Atomic:
    • If a transaction contains more than one transaction, they are grouped as one unit: If one operation fails, the whole transaction fails.
  • Consistency:
    • There is no stale state in the database. Everything is kept up to date.
  • Isolation:
    • Each Operation is done in the database as if sequential.
  • Durability:
    • The effects of an Operation is stored on disk forever.

• Indexing:

  • Indexes allow for fast reads of a table:
    • if you have to find one record on an unsorted table, that is an O(n) operation.
    • If we index it (restructure the table as a tree), this can take O(log n) time.
    • They do cost extra space, and increase memory usage.

• Remember to always write queries as transactions:

BEGIN TRANSACTION;
 
-- ROLLBACK if condition is not met
- ROLLBACK;
 
COMMIT;

Effectively, SQL is run sequentially.

Prev: [hashing](hashing.md) Next: [key-value-stores](key-value-stores.md)