The wonderful thing about HyperLogLogs is that you can split the counter in N servers and "merge" the registers later, in case you want an architecture that sha

The wonderful thing about HyperLogLogs is that you can split the counter in N servers and "merge" the registers later, in case you want an architecture that shards the same counter in multiple servers. But sharding directly by resource looks simpler actually...