Edgeryders test customers for Jitsi servers

@nadia @owen @matthias

Owens solutions works great for quick and video calls, but it is not a great solution if you need:

  • To schedule and invite to a call
  • Securely protect a call and only invite certain people
  • Be sure that the call is end not listened in on
  • Allow for host control of muting participants or kicking out if needed
  • Scale to calls of 20+ people

At the same time, my friend Peter Tapio just started working on an offer to set up dedicated Jitsi servers for companies. When we tried Jitsi, we tried their free “Jitsi Meet” service, which has horrible latency. A dedicated server would work a lot better. Peter is offering us to be his test subjects for a month. All we pay for in the first month is the server costs of 50 USD per month.

A very interesting benefit of this is that Jitsi can use the Edgeryders Discorse as its single sign on server. You would be signing in to the video client with your Edgeryders account. Anyone can create a room on demand. We are no longer dependent on sharing the logins for some Zoom account.

I asked him for a quote.

We can easily try out configurations and test for a day with very low cost (that I would not bill).

If we get it working, I’m thinking around 50€ for the month depending on the server resources. After a month of testing we can see the bandwidth costs and all. It should fit within the budget, unless there is an unusually large amount of calls happening. We can even scale the server if we need more resources. Testing this should be flexible. The included bandwidth is ok for over 100 hours of 5 people chatting with video by my estimate. Extra bandwidth after that is 0.02€/Gig. The server is on a 500MB link, which would presumably support about 100 concurrent users. But this can be limited by other resources at some point as well.

Looks like all authenticated users can create rooms and usually anyone can join. You can password protect the rooms (individually on setup) if needed. There is also an option to stream a room to YouTube if that is interesting.

If that works well, we can continue with a reduced rate and I’ll start selling it to others more. I’m offering this initially for the quarantine times, to supplement my income. Depending on how well the business goes, I may continue afterwards.

Should we try this? It seems like at 50 USD, it’s a very worthwhile experiment.

1 Like

i say yes. the sooner we can ditch zoom the better.

Cool, I would like at least the go-ahead from @matthias too since it may require a little of his time to help Peter connect to our Discourse to authenticate.

I have never been in a single Jitsi call that deserved the name. Mostly I could barely understand anything.

Can he offer a video conferencing environment that we can test out with 15 people without setup efforts? I would not want to waste a day of setup work to find out it just does not work :stuck_out_tongue:

1 Like

Were these calls on dedicated servers or the free Jitsi Meet servers?

Yes, he is offering to set it up without any fees for us to test before paying anything. I guess we can do that first without going through the authentication with Discourse steps.

1 Like

It sounds to me more that we’d have to work for a day to get something working? That’s what I want to avoid. It also sounds like he’s still testing out the best way to configure a Jitsi server and is rather at the beginning of offering this as a product? That’s no deal-breaker, but … I got so annoyed with Jitsi that the first thing I need to see now is that it can be used to make a call.

(And I don’t know if the Jitsi “calls” I experienced were on Jitsi Meet or not … if the software is not able to detect bandwidth limitations and take corrective measures, there’s probably something missing in the software …)

No, I don’t think we need to do much. I think he will set it up and we just bring people and try it out. At worst, the time we waste may be trying out having one of our community calls break down, and we can always have a Zoom call as a backup.

But I only want to take the offer if we are in fact ready to pay for it in case we like it. In that case, it would require work on our end for the authentication through Discourse.

Would be happy to pay for open source infrastructure that works. The solution Owen is working on is as little open source as Zoom when it comes to the actual video component from daily.co.

If the initial call is promising, I’d be fine with signing up for a month and testing further. If it works, Owen can hopefully exchange the daily.co component with a Jitsi component, and we integrate the Discourse authentication in that way.

In other words, I’d not want to work on the Discourse integration until our first month of more extensive tests is over.

3 Likes
  • To schedule and invite to a call
  • Securely protect a call and only invite certain people
  • Be sure that the call is end not listened in on

As of the latest update, these are no longer missing features… You can create a call by referencing the id of the call topic, for example: https://now.edgeryders.eu/call/12876

This will only allow a user to join the call at the scheduled time, and only for the users listed in the RSVP. You can also limit access by discourse groups.

For the other two points, muting and kicking out participants is very easy to implement with the Daily.co API. I haven’t been able to test it with 20+ people but from my experience with it, the quality and latency is as good or better than any other service I have tested.

This isn’t to say another solution wouldn’t also work - but I would consider carefully where our resources should be spent if we have something working already…

I think the shortcomings you listed are less related to the API we are using, as they are to UI features that can be implemented… but there isn’t much sense in implementing them if we’re not sure which solution to go with.

From what I read about Jitsi, the benefits would be that it is open source, which I understand aligns better with the core values (but then again we are still using services like YouTube and Mailchimp, which are proprietary…). But the downside would mean running an extra server (just talked to @matthias about how we would like to avoid this), maintenance and associated costs - for something that does one thing.

The admin interface would also be an extra level of UI, arbitrarily disconnected from our platform - while the current system relies quite nicely on topics on the platform - that are open and editable.

This isn’t merely to play devil’s advocate (or push us for using daily.co - of which I own no stock :slight_smile: ), but just to point out some things that should be considered before jumping into another pond…

/cc @nadia @hugi

3 Likes

So I have been running Jitsi Meet servers for some time. The test part is where you would need to configure your platform to use the JWT plugin for auth. It should be mainly an issue of coordinating the settings.

Jitsi meet servers are in the US, and the videobridge routes all data, so that may be why Jitsi Meet is not eh best in EU. I have an instance running in NL if you wan to test?
:Ptr

You could try BigBlueButton - I was in a call with 10 ppl last Saturday that worked quite well. It’s meant to be for online classes so probably works for even more people, with some caveats for distance etc.

1 Like

Here is the discourse plugin: GitHub - discourse/discourse-jwt: Discourse Auth support for JSON Web Tokens (JWT)

With that and a shared secret Meet can auth users from dicourse.

1 Like

Worried this could be getting more complicated than necessary…

I suggest we test it with an already running instance (maybe the one that @notme mentioned (thanks :slight_smile: ) and see if there are any real advantages to using it over what we have.

1 Like

Not sure if the bandwidth will last on the current one, if you want to test for a month?

For a full month of testing I can build a server for you. I can handle everything if you cover the cost (50€ for dual core). We can also try the JWT auth, as it only requires config on my end… If you can get JWT token auth to work on your side.

What would be a good date for the free one day test? I’ll invite others too so we can stress test the server. We can combine testing capacity if the time is suitable for all.

I think @hugi will coordinate the test day since he proposed the migration to Jitsi.

Would we have to install the discouse-jwt plugin already for the test or not? It would be kinda ok to do so, but the less work to see a demo of Jitsi with 20 people, the better. After a successful demo, we’ll be happy to invest some time into integration of course.

@MariaEuler, @nadia, @noemi
Do we have any community calls in the coming week that we could try this on? It would be interesting to get some people to do the first half of the call on @owen’s setup and then try Jitsi, so we can compare.

1 Like

How many people would you need to make it a vaiable test?

I think 20 would be great, but I think we could pull some more people in for just testing if they can just join the call an keep working like normal, no?

Was thinking asking the people from the open source call tomorrow evening to test it out. Could be a programm point there in itself. however, we would need to try to get some more people in there.