Skip to content

Efficiently handle sparse files in StorageBackend::get_into #527

@asomers

Description

@asomers

StorageBackend::get_into is the main way to read a file. Currently, it simply reads the entire file from the storage backend, and copies it to the Writer. But that's inefficient for sparse files. It would be better if it understood sparse files, and could synthesize zeros as needed. This would require:

  • Adding an lseek method to the trait
  • Implementing the lseek method for the FS backend
  • Within get_into, iterating through the file using lseek with SEEK_HOLE and SEEK_DATA
    • Copying data from the storage backend for every data region
    • Synthesizing zeros for every hole region

Alternatively, this change could be made entirely within the FS backend by returning a custom object that implements tokio::io::AsyncRead but is hole-aware, and synthesizes zeros as needed.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions