Motivation

The current Apache Ignite Storage implementation has strong coupling to the B+ tree storage. This means that we can't add another implementation of it. The B+ tree is useful for one of the types of loading profile and usage.

If we get rid of a strong coupling, it gives the following advantages:

1. The b+ tree storage will be easier for testing.
2. This is an opportunity to improve performance benchmarking.
3. B + tree can be provided as independent project / module and be reused somewhere.
4. This is an option to add another storage:

4.1 We can add other types of storage for different types of tasks, load profile, hardware such as column-oriented storage, LSM, and so on.

4.2 This is an opportunity to add a more stable solution that is fixed and tested by other open source communities.

Description

Discussion Links

Tickets