Hi,
we are considering use wowza. (RTMP/HLS)
but the concern is the architecture.
how can I achieve multiple origin and edge? is it possible through native WSE feature?
I already saw the some of the docs (like live stream repeater, load balancing…), but I couldn’t find it.
our workload is like twitch (small scale), and I want to achive auto scaling for origin and edge.
could you advise me?
Thanks.
Welcome to the Wowza forums. So if I understand you correctly you want to be able to scale from multiple Wowza servers configured as the origin?
You could do this but you would have to pay for more than one Wowza server instance. Have you considered doing this in Wowza Video instead where it will auto scale for you?
If you’re trying to do something like Twitch where multiple people can upload a video to your app, you can do this through either Engine or Wowza Video.
I would reach out and talk to to a sales engineer to find out which one would make the most sense for you. I can share some info on scaling with multiple Engine instances for now.
http://community.wowza.com/t/is-it-possible-to-run-multiple-instances-of-wowza-service-on-1-machine/54231/2
wowza does not support out-of-the-box autoscaling for WSE, which is why the Wowza cloud exists. That said the articles you mentioned are building blocks toward multiple origin/edge scenarios.
-
Live Stream Repeater : This is the starting point. The arrangement of origin & edges in a streaming architecture is called a cluster. Live stream repeater will help with scaling a stream from one origin to multiple edges. Wowza docs used to have everything in plain configs earlier. Now it is just clicks and buttons. so for in-depth knowledge automation, you need customer support.
-
Load Balancing: This will be applicable when you want to distribute traffic evenly. this is applicable to both ingest and egress. A load balancer is only effective when it is aware of the real-time load of the cluster. Therefore needless to say Wowza load balancer will help you more than any round-robin or third-party LB. Unless you gather real-time load data you cannot distribute the load fairly. You need to know which nodes are free and which nodes are heavily loaded.
Then of course there is geo load balancing. but that is a different story to discuss.
- Autoscaling : Now coming to autoscaling. Autoscaling is a function of both the above activities mentioned. i have been looking for some options on wowza too but couldn’t find any. the theory would be like this. CPU load-based autoscaling is still possible on most clouds but you need to take care of configuring new instances as origin/edge dynamically. Lots of streaming products use this system as it is an out-of-the-box offering.
The other part of this story is scaling by custom metrics. Again that is a different beast altogether. It depends on multiple factors and specifically the nature of the specific product.
So all in all…as @Rose_Power-Wowza_Com suggested, you should start at Wowza sales to understand how you can read your goals faster. Custom autoscaling solution on Wowza cloud.