Option 1: Set up a sharded cluster to distribute writes. However a sharded cluster brings complexity in terms of additional components needed (mongos, config servers, multiple replica sets, etc), along with technological overhead you'll need to deal with (balancer, chunk migration, no point-in-time backup/restore, etc).
Option 2: Set default writeConcern to 0, a.k.a. 'fire and forget' mode. Read more here. Basically, your application writes to mongod process (mongod in Primary, if you use a replica set), and receive no acknowledgement back. Two ways to use writeConcern:
- Set as default across replica set
- Use writeConcern per query
Read more about it here.