Show HN: Triton, a scaleable media center | Hacker News Reader

Show HN: Triton, a scaleable media center

github.com

13 pointsjaredallard13 comments

I read your "What is TRITON?", and didn't know what TRITON is afterwards. So read the rest of the doc. Still don't.

What will I see as an end user? An actual video player for playing my video library? You compare yourself to Plex at one point, and I know that's what Plex gives people... Or is it just some sort of API end-point which you need to somehow attach a video player to?

So it will be just like Plex but it will be more focused on larger media servers or people who want to not care about organizing media, need multiple qualities pre transocded, and etc. I wrote a bigger detail to inimino below about why that decision was made, but it's s3 first class and all about ability to server media globally and scale.

Hi everyone, I'm the lead maintainer and original creator of this project. I'm interested in hearing feedback on my project, which is a microservice oriented alternative to Emby/Jellyfin/Plex/etc that focuses on being scaleable and being cheap to host in the cloud. It solves the ideal use case of being a high load media provider, focusing on runtime costs (encode once, not every time) and being able to handle, of course, high load. Running a media server can be a massive pain, and being able to abstract a lot of that away (autodetecting media names and episode numbers) has saved me a lot of time for my personal collection. Interested to hear anyones thoughts!

Might consider a different name since Triton is already a cloud platform from Joyent.

https://www.joyent.com/triton/compute

Yeah, probably should.

On the other hand, if you could get Triton running on Triton, it'd mean I could run a media server off the SmartOS box I've got at home (or the multiple SmartOS boxen I've got at work) :)

You include usenet in your example of possible video sources; it is my understanding that cloud providers will ban your account if they get DMCA requests. I am not certain how much S3 providers monitor storage for this type of thing, but I would guess they do up to a point. Do you think this could be a problem?

Yeah, after a few DMCA requests they will ban your account. Usenet is supported and torrents, but that is not intended to be used for copyrighted material. AFAIK s3 doesn't do any scanning, nor would they care, since DMCA safe harbor protects them until they get a notice. There are plenty of options if you do decide to be a pirate to stay safe, but again, I do not encourage or think it's a good idea.

> Emby/Jellyfin/Plex/etc

I don't know any of those things. Who is this for? Why would I use it? Why is it better? Why is it using fancy shit like k8s for no apparent reason? What is "high" load? Why are runtime costs important?

Good points, I should be more descriptive. Jellyfin and co are media centers, they provide the ability to stream media from a hard drive with a nice user interface.

This is for someone who wants to run their media server w/ a lot of users, or someone who wants to be able to run a single server and not have to deal with transcoding to different formats. As of right now, it's not "better" than these others since the frontend isn't finished, but when it's done it'll be better in two main ways; able to scale the frontend, able to reduce global latency by using s3 bucket(s), able to have a full e2e approach to media importing (everything is the same format, same subtitles, same layout by default) usually that has to be done manually by whomever operates a server. It's using k8s so that individual microservices can be scaled using the autoscaler microservice. This allows users to use premptible/spot machine types in their workloads and only pay when they are using them, which makes converting cheap.

High load is users watching a lot of media, to get S3 to work w/ any alternatives you have to run gcsfuse which is expensive to use and a bottleneck when even two people watch different shows. Normly you'd run two or more instances then to load balance, well, with all of these they use sqlite so there is no load balancing capabilities.

Runtime costs are important to anyone who has tried to run a single Plex server in the cloud and had to pay a lot of money to be able to have fast, real-time, on demand transcoding so a user could watch a movie/tv show. Ideally you want to pre encoded all of those, that was the original goal of this project, support that. Now it's turned into a full e2e approach to remove dependency on those media centers that aren't cloud first.

I hope that explains it better, it's a little niche, but for anyone handling a lot of users I think it's useful, either way for me it's useful.

Thanks! I really appreciate the explanation. Hope some of this can make it into the README as well and be useful for other non-experts in this space :-)

What's the best way of following this project? I want to be kept up to date on progress. Is there an RSS feed or an announcement list?

I don't have anything currently, but I'll see what I can do about getting some sort of announcement list and will ping you. Ideas are welcome!

Reply on news.ycombinator.com