Skip to content

Use LMDB for storage #96

Description

@erikbern

@LongbinChen pointed out that LMDB would be a good way to support a series of things.

http://symas.com/mdb/microbench/

If we swap out the storage to rely in LMDB we would get a lot of features for free

  • Ability to use any key, not just integers
  • Out-of-core data (dataset doesn't fit in RAM)
  • Decouple index structure from nodes (could potentially set _K independently and optimize for it).
  • Incremental add_item (and deletes) would be much easier to implement

I'm not sure how tricky this would be, but probably not more than 5-10h. Mentioning it in case someone wants to give it a shot :)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions