Netflix divides its infrastructure into two distinct components, commonly described as a cloud-based "brain" and thousands of local storage units known as "movie lockers".
This split challenges the common assumption that streaming video travels directly from one enormous, centralised server.
What happens when you press play
When a user starts a video, the request first goes to the cloud brain, which is hosted on Amazon Web Services.
This cloud brain performs several administrative tasks before any video begins.
It verifies the user's subscription, works out the video resolution suited to their device, and identifies the storage server sitting geographically closest to them.
Where the video actually comes from
Crucially, the cloud brain never sends the video file itself to the user.
Instead, it hands the user's device the address of a physical storage box inside their own internet provider's facility.
The device then connects directly to that nearby box to stream the content, bypassing the cloud entirely for the heavy lifting.
Keeping track while you watch
While a user is watching, their device sends small, periodic updates back to the cloud brain.
These updates record playback progress, including the exact point at which a viewer pauses.
Why the split matters
Separating decision-making from video delivery is the key to the system's efficiency.
Netflix calls the decision-making side the "control plane" and the delivery side "Open Connect".
This architecture allows millions of people to stream simultaneously without overwhelming the wider internet.
A global network of boxes
The Open Connect network consists of content delivery nodes placed in locations around the world.
These include cities such as Manhattan, London, Frankfurt, Mumbai and Tokyo.
By storing popular content close to where people actually watch it, Netflix keeps videos loading quickly while avoiding the congestion that a single centralised system would cause.