The Fastest Database in the World?
If you ask a low-level engineer what's the fastest database, the honest answer is: pure data structures.
All traditional databases are just data structures + extra logic (overhead). Speed always comes down to CPU instructions per operation and how well they fit the hardware pipeline.
Example: ClickHouse Simplicity
ClickHouse started with two consoles: one constantly writes data to disk, the other merges files. Simple, right? Any DB that adds unnecessary complexity will lose performance—especially on write operations.
What is Storage in Fractal?
Fractal doesn't use a traditional DBMS. Instead, it introduces the concept of a Storage — a minimal key-value engine, similar to what some DBs call an "Engine".
A working Storage can be built with just 300–400 lines of code. Each one can be specialized for:
- High-speed writing
- Fast reads
- Small documents
- Large blob files
- FIFO queues
- Compression & encryption
- Row- or column-based formats
All of these Storage types can live under one application — a true hybrid DB inside your app.
Deep Granularity
In traditional databases, granularity often stops at the table or collection level. In Fractal, Storage can be as granular as a single field in a document.
Do You Need to Understand All of This?
Not really. Just like in other databases, 95% of use cases are covered by the default, balanced Storage.
But when you're in that 5% of edge cases — with massive data or performance needs — you’ll have full control and flexibility.
How Fast Can It Get?
In synthetic benchmarks, specialized Storages can handle 3–5 million operations/sec or more. They can even process petabytes of data like ClickHouse-style systems.
Why? Because Storage is just simple, low-level algorithmic logic.
Dimensions: Optional Features
Things like Transactions, Undo, and Patch mechanics are implemented as Dimensions — modular layers you can turn on or off.
You decide how close you want to get to the theoretical hardware maximum.
Need to Use Relational Databases?
No problem — a Storage can act as a data adapter to MS SQL or PostgreSQL.
Typical setups: Storage (master) → PostgreSQL (replica), or PostgreSQL/MSSQL as the primary data source.
Switching from document-based to relational storage is easy — Fractal can automatically convert any JSON document into normalized relational tables.