Thursday, December 01, 2016

Implementing Automatic Digital Preservation for a Mass Digitization Workflow

Implementing Automatic Digital Preservation for a Mass Digitization Workflow. Henrike Berthold, Andreas Romeyke, Jörg Sachse.  Short paper, iPres 2016.  (Proceedings p. 54-56 / PDF p. 28-29). 
     This short paper describes their preservation workflow for digitized documents and the in-house mass digitization workflow, based on the Kitodo software, and the three major challenges encountered.
  1. validating and checking the target file format and the constraints to it,
  2. handling updates of d content already submitted to the preservation system, 
  3. checking the integrity of all archived data in an affordable way
They produce several million scans a year and preserve these digital documents in their Rosetta based archive which is complemented by a submission application for pre-ingest processing, an access application that prepares the preserved master data for reuse, and a storage layer that ensures the existence of three redundant copies of the data in the permanent storage and a backup of data in the processing and operational storage. They have customized Rosetta operations with plugins they developed.  In the workflow, the data format of each file is identified, validated and technical metadata are extracted. AIPS are added to the permanent storage (disk and LTO tapes). The storage layer, which uses hierarchical storage management, creates two more copies and manages them.

To ensure robustness, only single page, uncompressed TIFF files are accepted. They use the open-source tool checkit-tiff to check files against a specified configuration. To deal with AIP updates, files can be submitted multiple times: the first time is an ingest, all transfers after that are updates. Rosetta ingest functions can add, delete, or replace a file. Rosetta can also manage multiple versions of an AIP, so older versions of digital objects remain accessible for users.

They manage three copies of the data, which totals 120 TBs. An integrity check of all digital documents, including the three copies, is not feasible due to the time that is required to read all data from tape storage and check them. So to get reliable results without checking all data in the archive they use two different methods:

  • Sample Method Integrity 1% sample of archival copies is checked yearly 
  • Specified fixed bit pattern workflow that is checked quarterly.

Their current challenges are in developing new media types (digital video, audio, photographs and pdf documents), unified pre-ingest processing, and automation of processes (e.g. to perform tests of new software versions).

No comments: