2 ms·
That would be my point if I was about to make a more elaborate post - I don't expect this to run 20ktps on a geographically distributed cluster, but if it can d
by baq 4y ago
That would be my point if I was about to make a more elaborate post - I don't expect this to run 20ktps on a geographically distributed cluster, but if it can do 100tps with synchronous replication between east and west coast it opens a different set of possibilities than 5tps. If it can do 20ktps... it makes my life a whole lot easier, which is quite fine by me, too ;)
- azurelake 4y agoYou can already do this with MySQL/Postgres/etc., and there's no special magic in any of these newer tools that allow you to avoid the speed of light. The reason people don't typically run WAN synchronous setups is exactly because of that (and network reliability). The only way to a "free lunch" is via CRDTs. edit: This is in no way a knock on FDB by the way. I think it's a really interesting piece of technology and I want to explore it more. Just that it doesn't have a way to do consensus faster than the speed of light :)