Drupal Making News

How do we handle 5 major Danish news sites, 250 online journalists, 5 million daily page views, 1 million news articles in the database and another 500 being added every day - all in one database?

Transcript

1. Trellon 25 Aug 16:00 Drupal Making News Arne Jørgensen & Christian Schmidt
2. Agenda • About Berlingske Media • Fundamental architechtural decisions • News articles as nodes • Caching • Questions
3. About us • Arne Jørgensen (arnested) Systems architect • Christian Schmidt (c960657) Systems architect
4. About Berlingske Media • Largest newspaper publishing group in Denmark • Established in 1749 by Ernst Henrich Berling • 62 print titles, including 11 national and regional dailies • 75 websites o 6 million users per month (unique cookies) o 175 million page views per month
5. BOND: Berlingske on Drupal • Project started in August 2008 • First site launched in May 2009 • Sites on BOND so far: o Newspapers: www.berlingske.dk, www.bt.dk, www.erhvervsbladet.dk o Verticals/sub-brands: www.business.dk, www.fri.dk, www.sporten.dk o Community: www.løbnu.dk
6. Roles • Journalist: Write news, choose pictures • Editor of the day: Prioritize news • Site manager: Modify Panels configuration • Developer: Write modules and themes, setup Panels • Sysadm
7. Fundamental decisions • All sites in one database • No Views, no CCK • Lots of Panels • Reverse proxy caching • Dedicated backend for journalists • All nodes have one primary term
8. All sites in one database • Several settings.php files that define $conf['theme_default'], one $db_url • 2.4 million nodes, including 1.2 million news articles and 600,000 image nodes • 60,000 taxonomy terms • MySQL InnoDB with transaction support • Master/slave database replication • Unpleasant surprises …
9. Nodes • Simple model • Primary term • URL alias • Presentation tags (terms) • Node references • Node listings: By time, or node queues
10. Primary term • Used for most node listings, sections etc. • Determines page layout for node view • Determines whether individual Panels panes are visible • Determines URL alias of nodes: www.sitename.dk/section/node-title
11. Caching • Memcache – take load off database • Reverse proxy – take load off database
12. Next step • Migrate regional sites to BOND • New caching strategy for logged in users • ?
13. Questions?
14. http://cph2010.drupal.org/node/14248