Skip to content

Implement in-memory on-write Chunk compaction #6812

@teh-cmc

Description

@teh-cmc

Chunks need to be compacted with their neighbors when they get inserted into a ChunkStore, so that all the runtime arrow downcast/introspection-related fixed costs can be amortized.

This is a pre-requirement to drop the deserialization cache, which is a requirement for megahertz logging.

There an infinite amount of subtlety to this, but for now we only want to focus on the very happy path (i.e.: make logging scalars nice).

Metadata

Metadata

Assignees

Labels

Type

No type

Projects

No projects

Relationships

None yet

Development

No branches or pull requests

Issue actions