4 ms·
The premise for MPEG-DASH is that infrastructure is easy: Just small files transferred over HTTP. You can use every CDN which supports HTTP for that. Live enco
by Jakob 9y ago
The premise for MPEG-DASH is that infrastructure is easy: Just small files transferred over HTTP. You can use every CDN which supports HTTP for that.
Live encoding is surprisingly hard: You have to get a live input signal instead of a file and transcode it on-the-fly to DASH with minimal latency meaning:
- encoders in the vicinity of the input signal,
- multiple input signals,
- efficiently transport a high data rate from a venue (with often a subpar connection) to your encoders,
- support many different input signal formats (i.e. camera output signals),
- support very high data rates etc.
Some companies do it to a varying degree but not that many like Wowza, Elemental (bought by AWS and put into AWS Elemental Delta), Adobe, Bitmovin, …).
- nik736 9y agoYes, but they are putting together a huge sales team, their customers are using their video player and a video encoding service. That means they also have the need for video storage, a video CDN, etc. So they waste a lot of potential. Building a encoding service in the cloud era, even for live video, isn't really that hard. The hard part is that every customer has different needs and different source files, if you build it inhouse you are only building it for one customer, yourself.
- mv4 9y agoCommodity storage and delivery prices have been declining steadily. Building something that is scalable but also profitable, is hard. Still, selling is the hardest part, in my opinion. Source: run a global video cloud.
- nik736 9y agoSo you are exactly confirming what I am saying, they have or are building the sales team anyways.
- mv4 9y agoOne thing I didn't make clear, that may explain why somebody just wouldn't build their own storage/delivery: when certain services become a commodity, and long-term commitments become an exception rather than the norm - it's easier to shift the pricing pressure and uncertainty for the commodity side to your providers. You specifically don't want to "build your own" in a market like that, but instead focus on specialized, high-margin services. One exception: when you can predict/control the demand (e.g. you are a Netflix), then you build your own, and still leverage providers for capacity on tap.
- robterrell 9y agoThe one part of live encoding that needs innovating is RTMP for ingest. It's crazy to me that this lousy protocol is the first step for most live video.
- kyriakos 9y agoYou can already use webrtc for that.