Skip to content

IBLT needs a revision to meet its space constraints #83

@novakboskov

Description

@novakboskov

This may look good enough at first glance but it is not optimal.

For the given decode success rate and maximum number of symmetric differences, there are optimal k (number of hash functions) and h (hedge factor).

For the maximum number of symmetric differences up to 1000 and the decode success rate above 99.5%, optimal k and h can be found here.
Space-optimal k and h are obtained via a brute force search. Apparently, there is no analytical result that can meet the strict practical requirements.

This is how the parameters are used in IBLT construction.

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions