When you press play on a series, tens of megabytes per second start moving across your screen without you lifting a finger. But behind that gesture there is a chain of processes we rarely stop to look at: codecs, content delivery networks, adaptive protocols and encryption. This is what happens, and how it really works.
From the studio to your screen: the journey of the bytes
Streaming does not send the complete file before playing it, but rather in segments: small chunks of between two and ten seconds of video that are downloaded and played almost in real time. This model is called progressive download and has become the standard thanks to two protocols: HLS (HTTP Live Streaming) and DASH (Dynamic Adaptive Streaming over HTTP). Both chop the video into pieces and serve the fragments over the same HTTP protocol used by any web page, which enormously simplifies the infrastructure: no specialized server is needed, a static one that knows how to serve files is enough.
The codec: compressing without it being noticeable
Raw video takes up absurd amounts of data: one minute of uncompressed 4K can exceed a gigabyte. For streaming to be viable, compression is needed, and that is where codecs come in. A codec (from the English coder-decoder) analyzes each frame and discards information the human eye barely perceives, while also exploiting redundancies: in a static scene, the background barely changes from one frame to the next, so only the difference is encoded. The current standards are H.264/AVC (the most compatible), HEVC/H.265 (delivers almost twice the efficiency) and the promising AV1, an open and royalty-free codec that uses artificial intelligence techniques such as block prediction through learning to compress even more while maintaining quality.
CDN: bringing the video closer to your home
Serving all those bytes from a single server would be a latency disaster. That is why platforms use a CDN (Content Delivery Network): a network of servers spread across the planet that store cached copies of the content. When you request a series, the system directs you to the geographically closest node through intelligent DNS resolution, so the packets travel a shorter distance. In addition, routing algorithms choose the least congested path at each moment, which is known as anycast.
ABR: the video that adapts to your connection
Your connection is not stable: it can fluctuate depending on the Wi-Fi, the time of day or the saturation of your provider. To handle this, streaming uses ABR (Adaptive Bitrate). The same content is encoded in several qualities (for example 480p, 720p, 1080p and 4K) and divided into segments identical in time. The player constantly measures the available bandwidth and the amount of data in the buffer, and decides in real time which quality to download next. If the network degrades, it lowers the quality to avoid interruptions; if it improves, it raises it. All this logic lives in the client, which in each request communicates which quality it wants to move to, in an exchange regulated by the MPEG-DASH protocol through manifest files that describe the available versions.
TCP, QUIC and live streaming
Most on-demand streaming travels over TCP (Transmission Control Protocol), which guarantees that each packet arrives in order and without loss. But TCP retransmits lost data, and live there is no time to retransmit: you arrive late or you do not arrive. That is why live broadcasts increasingly use QUIC, a protocol built on UDP that reduces connection latency and allows multiplexing several streams without a slow one blocking the others. Platforms such as sports services and cloud gaming depend on this low latency so the viewer sees what is happening at the event almost at the same instant.
DRM: the encryption that protects the content
Not everything is served open: a large part of the catalogue is protected with DRM (Digital Rights Management), an encryption system that prevents copying the video. The segment travels encrypted and only the authorized player decrypts it using keys obtained through a secure exchange with a license server. The dominant standards are Widevine (Google), FairPlay (Apple) and PlayReady (Microsoft), and the security level is measured on a scale called L1 (hardware decryption, the most secure) or L3 (software decryption, more vulnerable).
Next time a series starts instantly, remember: it is not magic, it is an orchestra of codecs, cache networks, adaptive protocols and encryption coordinating in milliseconds. And on the broadband of the future, with the new codecs and QUIC, that start will be even more immediate.
