Tarsnap - Online backups for the truly paranoid

Navigation menu

What metadata does Tarsnap store?

Normal user data is split into chunks which are (on average) ~64 kB before compression. The actual length of chunks is variable so that if a file is modified in the middle, the chunk boundaries will re-align quickly (on average after 1.6 chunks) so that existing chunks can be re-used.

Tarsnap stores the following metadata:

  • A 512-byte tar header for each file. After these headers are passed through the deduplication and compression layers, each header is ~50-100 bytes.
  • Lists of ~1600 data blocks (i.e. which blocks belong to which files). These lists are also deduplicated and compressed.
  • A list of the lists of blocks, which is not deduplicated.
  • The archive name, time, command-line which created the archive, etc. This is approximately ~2 kB before compression.

Privacy note

All of the above data is encrypted with your private keys prevent anybody from infringing on your privacy — even the names of archive are protected.

Tarsnap keeps a local cache which keeps track of which blocks it has previously uploaded; querying the server for every block would be much too slow, and would reveal more information to the server about how archives are changing over time. The cache directory is ~0.5% of the size of the data you store.