Network subsystem overload detection?

Question

Xen

Good day. I want to use a xen-based cloud for video streaming, namely Selectel's. The load varies from zero to several thousand online viewers. I want to use the following scheme, one dedicated server to which streams are received, where they are packed into hds / hls, and several cloud servers used as a cache. So, the correspondence with technical support did not answer the question - how do I know that on this particular server instance I hit the network bandwidth ceiling? And I need to raise another server.
I plan to use nginx - as a proxy server, OS - debian with the Selectel kernel.

Answer 1

On a VPS, which is essentially the SELECTEL cloud, you will not know in any way except for the problems with video and packet loss that you have hit the network.
Nevertheless, the selectel promises up to 900Mbit per server with relatively no problems.
If I were you, I would go up to 500Mbps and then raise a new server, leaving a margin of safety for the current one, all the same, payment is only for the resources actually used, and in this case there will not be much difference in terms of money between 9 servers of 500Mbps and 5 of 900 ...

Answer 2

I would advise you to just test and detect the threshold at which the traffic begins to shape.
To be honest, I didn’t work specifically with Selectel, but I suspect that they cut the channel at certain limits, like most cloud VM sellers.
It's strange that tech support didn't say anything.

Answer 3

It is very unreasonable to host videos on a VPS, and even more so in the clouds.
First, it's frankly expensive.
Secondly, you will not get a normal channel, despite the assurances of sales, nor stable disk performance - both are shared with neighbors.
Thirdly, you practically lose the ability to monitor the load.
For your task, you need dedicated servers with dedicated channels and good disk arrays.
And about 500Mbit on the selector - this is a joke, apparently.