Link to video

Suppose you have a large pizza that you cannot finish by yourself, you divide the pizza into small parts so that you can share it with your friends. We need friends to finish the pizza, effectively we are taking the range of the pizza and breaking it into pieces. We "partition" the pizza.

<aside> 💡 Analogous to above example, when we use a key to break our data into pieces and allocate that to different servers, it is called horizontal partitioning.

</aside>

It depends on one key that is an attribute of the data we are storing. For ex : Use location as a sharding key.

<aside> 💡 Vertical partitioning is another way to partition data which partitions data according to columns.

</aside>

Some properties of data that are important:

• Consistency
• Availability

Some properties of sharding :

• Joins are very expensive as they are made over the network.
• Shards are inflexible. We can use hierarchical sharding to get rid of inflexibility.
• Creating indexes on shards are very helpful. They make queries fast. Sharding makes read and writes faster as all queries fall on one particular point.
• To handle failures in sharding, use master-slave architecture. All writes happen on master and meanwhile all slaves continuosly poll the master and read from it. If there is a read request, it can be distributed across the slaves. In case master fails, slaves choose one master among themselves.

Understanding Database Sharding | DigitalOcean

This is another great blog you can read to understand sharding in more depth.