Fedora 3: A Smooth Migration
MetadataShow full item record
The dramatic changes between the architecture of Fedora 2 and Fedora 3 offer exciting opportunities for improved functionality and organization through the addition of a more formalized content model architecture (CMA). On the other hand, these changes may make the migration of existing production repositories seem a daunting task, where one must balance the advantages of new features with the requirements to maintain uninterrupted and unaltered service. In this presentation, we will present the case of the 2008 migration of the Indiana University Digital Library Program's Fedora repository from version 2.2.4 to 3.1. The first portion of this talk will be dedicated to technical and logistical challenges associated with the migration of a repository of nearly half a million objects. These include balancing ingest scheduling with migration timing, dealing with a large amounts of data, and switching from Oracle to MySQL as the relational database system. There will be discussion of the considerations associated with migrating to the new CMA and use of the "generator" application that is part of Fedora 3's migration tools. The pros and cons of various techniques of modeling standard types of objects such as images, books, and serials will be presented as well as the advantages and disadvantages of our outcome. Finally, some time will be spent exploring the implications of migration to Fedora 3 on the custom tools and services that manage ingest, search and delivery in our repository. These include a Lucene-backed search service that was updated to work with the new messaging architecture, an ingest tool that had to accommodate new content models and a new version of FOXML, as well as our identifier resolution service used to maintain persistent URLs to objects (based around the OCLC PURL Resolver service).