After the recent problems with the Google Network - I think that Roon developers should try to improve that piece of software that handles network communication events (change buffering may be?). Just like developers of industrial automation systems dedicated for nondeterministic networks do. The assumption that every “Roon network” is deterministic could be wrong.
It could solve many other problems, that not seems like network problems…
I have noticed - there is no official statistics, but I have noticed that there is more service requests when the network slows down.
What was the recent problems with ‘Google Network’?
I don’t think we as consumers can expect any application developer to compensate for a poor network. Doing so would lead to compromises many won’t like.
It’s far easier to build or fix the network so it’s working properly in my opinion.
A few months ago, we hired a domain expert in this area, and now we are in the midst of a significant migration that will make Roon more robust to outages like that one, will make our backend more scalable to varying load throughout the day/week/year, will improve our internal monitoring/alerting tools, speed up disaster recovery, among many other benefits.
This project is well under way, but it’s not a quick effort–first new infrastructure needs to be developed, then dozens of interconnected services have to be moved from the old infrastructure to the new infrastructure one by one without breaking anything while simultaneously developing better operations processes for the team. I expect to see things incrementally getting better over the next 6-12 months, at which point we will be running completely on the new stuff.
Google had a major power outage at the data center that hosts our services, and everything was offline for hours. The standard way to mitigate this is by running services in multiple data centers for redundancy. This is operationally a lot more complex for us, and is also a cost multiplier. Nonetheless, we are transitioning to this model (see above), but it will take time.
Ah, I was blissfully unaware. Thanks Brian. I misunderstood the OP to be referring to local network connectivity.
I have many services I manage for the company I work for hosted in AWS but we use 3 regions and multiple availability zones in each region to mitigate the effects of a regional outage.
The additional layer of network/resource abstraction enabled by cloud hosts is a speciality no doubt. I encourage taking all the time you need to ‘get it right’ as undoing and redoing is costly in negative user experience.