🗓️ 31012025 1637
📎 #wip
TLDR
Row-oriented storage stores entire rows of a table sequentially on disk or memory
In this model, all columns belonging to a particular row are stored together
How It Works
| ID | Name | Age | Salary |
|---|---|---|---|
| 1 | Alice | 30 | 100k |
| 2 | Bob | 35 | 120k |
| 3 | Carol | 28 | 95k |
Table of data to be stored
[1, Alice, 30, 100k], [2, Bob, 35, 120k], [3, Carol, 28, 95k]
How data is stored on disk
Pros
- Efficient for oltp workloads
- Best suited for transactional systems where queries typically access one or a small number of rows at a time (e.g., retrieving user profiles or updating orders).
- Fast for point queries and updates:
Accessing or updating a specific row (e.g., based on a primary key) is faster because the data is stored contiguously.
Cons
-
Not ideal for analytics:
Analytics queries often access only a few columns (e.g., aggregating salaries). With row storage, reading the entire row introduces unnecessary I/O overhead. -
Larger disk usage for sparse data:
If many columns contain nulls or zeros, row-oriented storage can lead to inefficient space utilization compared to columnar storage.
Best Use Cases:
- OLTP workloads where frequent inserts, updates, or deletes are common (e.g., user transactions, bank systems).
- Point queries like "find employee details by ID."