Discussion: Backend architecture

[aeble]

considering the backend seems to time out at the moment it might be a good idea to redesign it. Are you interested in discussing it?

[KingUdo]

Hey @aeble

The Datacenter where we are hosting the backend system is having issues resulting in the first layer to become unresponsive.

Super happy to discuss how we would be able to run this decentralized in the future!

From what we heard the systems will be back up and running on Sunday...

[aeble]

I would suggest a tiered setup with more than one system to collect the data which in turn forward their aggregations to the backend that "does things" with it. It could be done as easily as creating more than one IP address for the api.netwatch.team name so it could be a round robin approach for the probes to load off their data and the aggregators either using the data or just pre-aggregating it so the backend could compute its tasks easier/faster. That way, one wouldn't need a load balancer for the time being.

When a probe can't offload its data it either tries again with a new DNS request or it stores the data locally to deliver it at a later point. [How to deal with DNS caches replying with the same IP address?]

If the aggregators need information from the backend that could be implemented as well.

One more question: do you have means in place that prevent the backend from malicious actors DDoS'ing it or, even worse, delivering false data to it? [Authenticating/encrypting/signing data with a local certificate that needs to be registered with the aggregators?]