3
edits
Changes
no edit summary
Keywords: Europeana, data quality, EDM, API, Apache Solr, MongoDB, #opendata, #openglam
== Teach your Fedora to Fly: scaling out your digital repository ==
*[mailto:acoburn@amherst.edu Aaron Coburn], Software Developer, Amherst College
*No previous C4L presentations
Fedora is a great repository system for managing large collections of digital objects, but what happens when a popular food magazine begins directing a large number of readers to a manuscript showing Emily Dickinson’s own recipe for doughnuts? While Fedora excels in its support of XML-based metadata, it doesn’t always perform well under a high volume of traffic. Nor is it especially tolerant of network or hardware failures.
This presentation will show how we are making heavy use of a fedora repository while at the same time insulating it almost entirely from any web traffic. Starting with a distributed web front-end built with Node.js, and caching most of the user-accessible content from Fedora in an elastic, fault-tolerant Riak (NoSQL) cluster, we have eliminated nearly all single points of failure in the system. It also means that our production system is spread across twelve separate servers, where asynchrony and Map-Reduce are king. And aside from being blazing fast, it is also entirely Hydra-compliant.
Furthermore, we will attempt to answer the question: if fedora crashes and the visitors to your site don’t notice, did it really fail?
[[:Category:Code4Lib2014]]