Blastr is very efficient. Both because #[4]​ did a great job with it but also because it only sends events.

Where you choose to host your relay and what implementation you choose to run will impact your costs. You can run a very lean private relay on a $10 VPS.

If you run a large relay, your bandwidth consumption will be substantial across all implementations. Picking a provider that allows for unmetered or has a high flat rate limit is imperative to control cost.

Taking it an extra step, if your users are going to be spread out geographically and performance matters, you’ll want to provide regional mirrors. Keeping these mirrors in sync and putting them behind a load balancer has additional costs.

The upside to cost doesn’t really have limits but there are low barriers to entry.

I’ve been most curious to do some work around note archival and perhaps curation. Specifically a concept that allows folks to save and publish/read events later regardless of relay retention policies.

My intuition coming into this question was that the cost for relay operators is a function of the number of events currently retained. As such, I wonder if non-relay apps could serve to free the relays up to have a shorter retention period and thus reduce costs while also providing some value to authors (especially of long form notes).

Or perhaps even a different style of relay would be more appropriate for this sort of thing.