Parse.ly: Inside a modern RIA built with Solr - Presentation Transcript
Inside a modern RIA powered by Solr
Andrew Montalenti
Co-Founder &
Technology Lead
andrew@cogtree.com
1
2
Mainstream
30,000
Blogs
1
900,000
1From Technorati’s 2008 State of the Blogosphere
3
But,
What about your interests? 4
your interests,
your web
5
6
What is ?
• Your unique interests
• … create a filtered, prioritized, and personalized news feed
• … built just for you!
• 120K+ news and blog sources tracked
• The most personally relevant items at the top
• Bottom line:
You spend less time skimming headlines,
and more time reading relevant content.
7
Demo!
(if possible)
Let’s pop open the hood!
9
The RIA
ExtJS + jQuery
JS REST Binding
REST API
django-piston
Solr Postgres
10
11
Per-Doc
Processing
Batch
size
IO-bound work
12
13
Solr in the Real World
• Storage of "canonical data“
• Relational vs. Search Index
• Complexity of custom relevancy scoring
• "Near-Real-Time" updates
• Solr in a pipeline
• Pushing bits and marshalling cost
• Index size, corruption, and stability
• Administrability
14
Scaling Up Parse.ly
• Custom scoring
• Multicore
• Distributed search
• Celery / Disco
• User-Article Binding Problem
16
17
Basic Almost There Comprehensive
solr.py
collective.solr haystack
pysolr
solrpy
solango python-solr
json/py output
18
0 comments
Post a comment