3 ms·
Looks like it's exponentially faster than RocksDB, but in most use cases the bottleneck would be the network and available sockets on the machine running FASTER
by welder 3y ago
Looks like it's exponentially faster than RocksDB, but in most use cases the bottleneck would be the network and available sockets on the machine running FASTER. Unless you're doing high throughput embedded data ingestion. Maybe Neuralink could use this.