ScyllaDB (written in C++)

ScyllaDB is a distributed wide-column database compatible with the Cassandra API.

Stores data internally using LSM Trees
noSQL databases scale horizontally due to LSM Trees, while SQL(BTree) donot scale horizontally
Super fast, it uses: Seaster Framework
C++ has predictable latency as no Garbage Collector Pauses wrt Cassandra(written in Java).

How ScyllaDB uses Seastar framework to store data?

Seastar framework binds(CPU affinity) 1 thread per CPU core with no IPC between threads making 1 thread as independent server(with its own RAM, disk etc)
Shard (in ScyllaDB terminology): An individual CPU core inside a single Node(Physical machine), along with the dedicated RAM and network thread tied to that core
When ScyllaDB starts on a server with 10 physical CPU cores, it launches 10 independent engine threads. It uses CPU affinity to lock (bind) Thread 0 to Core 0, Thread 1 to Core 1, and so on.

|----------------------------Physical Machine-------------------------------------|
|                                                                                 |
| RAM |0-25|               |26-50|            |51-75|             |76-100|        |
|       /\                    /\                /\                  /\            |
| |---- Core 0 -----| |---- Core 1 -----| |---- Core 2 -----| |---- Core 3 -----| |
| | Thread 0        | | Thread 1        | | Thread 2        | | Thread 3        | |
| | Hashes ending 00| | Hashes ending 01| | Hashes ending 02| | Hashes ending 03| |
| |-----------------| |-----------------| |-----------------| |-----------------| |
|                                                                                 |
|---------------------------------------------------------------------------------|

user_id = "Alice" --> |Hash| = 1042 -> 1042 % 4 = 2 (Shard 2)

Thread 2 (pinned to Core 2) processes and writes Alice's location data into Core 2's dedicated RAM buffer.