Crystal Clear: DeHub Stream App with 4K Support & High Uptime
By DeHub Team — 2022-04-01
The DeHub stream app shipped in Q2 2022 with 4K support and 99.9% uptime.
Two numbers, and they are not equally hard.
4K is table stakes
Supporting 4K is largely a matter of doing the encoding work properly and paying for the bandwidth. It is expensive rather than difficult, and by 2022 a platform that could not do it was not a serious option for creators who had spent money on their setup.
What made it worth doing was less about resolution than about not being the reason somebody's production looked worse than it was. A creator with a good camera, good lighting and a fast connection should not have their work degraded by the platform carrying it. Anything less makes you the amateur option regardless of the rest of the product.
Uptime is the hard number
99.9% is roughly eight and a half hours of downtime a year. That sounds generous until you consider where those hours land.
Streaming failure is not like a website being slow. A creator who goes live to a scheduled audience and cannot broadcast has lost that audience — not deferred it. The people who showed up have gone elsewhere, and the creator has to spend goodwill explaining a failure that was not theirs. Do it twice and they stop scheduling on your platform.
That asymmetry drove essentially every architectural decision in the stream app. The question was never "how good can this look" but "what happens when a component fails mid-broadcast", which is a much less enjoyable thing to design around and considerably more important.
What that meant in practice
- Ingest and playback fail separately. A problem serving viewers should never terminate the broadcast, because a recoverable viewing problem is survivable and a dropped stream is not.
- Degrade before dropping. Falling to a lower bitrate is an inconvenience. Ending the stream is an incident.
- Redundancy where a single failure would be visible, which is a different and much shorter list than redundancy everywhere.
The cost problem underneath
None of this is free, and 4K makes it worse. Streaming economics are brutal in a way that is easy to miss from the outside: the more successful a stream is, the more it costs, and unlike storage those costs do not amortise.
That pressure produced the advanced video compression work in the same quarter, which cut bandwidth costs by around 60% — not a nice-to-have but the thing that made offering 4K at all financially survivable.
Where it went
The 2022 stream app is the direct ancestor of everything since: Livepeer integration for 50,000+ concurrent viewers, a custom CDN at sub-200ms globally, and 99.99% uptime three years later — an extra nine that took considerably more work than the first three.
The standard set here has not moved. A creator's stream failing is our failure, regardless of cause.
Why "regardless of cause"
Because from the creator's side the distinction does not exist.
A stream that fails because of a provider outage, a bad deploy, a regional network problem or a bug in our code all look identical to the person who scheduled it and to the audience that turned up. Attributing it externally is accurate and completely useless — the creator still lost the session, and still has to explain it.
Adopting that framing has a practical consequence we did not fully anticipate. It rules out a whole category of comfortable engineering decisions, because "the provider was down" stops being an acceptable end state and becomes a design requirement to route around.
That is a large part of why distributing delivery eventually became a priority rather than a philosophical preference.