Hi, my name is Vladislav, and I am a member of the Tarantool
development team. Tarantool is a DBMS and an application server all in one. Today I am going to tell the story of how we implemented horizontal scaling in Tarantool by means of the VShard
Some basic knowledge first.
There are two types of scaling: horizontal and vertical. And there are two types of horizontal scaling: replication and sharding. Replication ensures computational scaling whereas sharding is used for data scaling.
Sharding is also subdivided into two types: range-based sharding and hash-based sharding.
Range-based sharding implies that some shard key is computed for each cluster record. The shard keys are projected onto a straight line that is separated into ranges and allocated to different physical nodes.
Hash-based sharding is less complicated: a hash function is calculated for each record in a cluster; records with the same hash function are allocated to the same physical node.
I will focus on horizontal scaling using hash-based sharding.