Skip to content
BENIBIX Learn

Separation of Storage and Compute

Also known as Decoupled Storage and Compute, Storage-Compute Separation

Data and the machines that read it are kept apart. Each can grow on its own.

Think of it like

It is like a library. More people can come in to read without more shelves.

Example

A team runs a small cluster all day and starts a big one for an hour each night. Both read the same tables, and nothing is copied.

Why it matters

You stop paying for machines that sit idle. Many teams can also use the same data at once.

Definition

Separation of storage and compute means the data and the machines that query it are kept apart. The data sits in one shared, cheap store. Groups of machines start when queries need them and stop when they do not. Each side grows, and is paid for, on its own. Older warehouses tied the two together inside the same servers.
On the right, compute can be added or turned off while the data stays put.

Common mistake

Leaving compute running when no one uses it. The data is cheap to keep, but idle machines still cost money.

Key takeaways

  1. 1
    One shared store. All the data sits in one place.
  2. 2
    Start and stop. Machines run only when needed.
  3. 3
    Pay for each. Storage and compute are billed apart.