The conventional wisdom of HTTP web page request between browser and servers is to transmit the response as whole in
the form of a structured mark up language. However, this may not be how it
works in today’s social networks.
Today’s smart social networks like Facebook use Hadoop and
Hive driven intelligence to predict which resources of web page have a
predetermined likelihood to be included in a response to a future request.
Resources which includes java-scripts, style sheets, image etc are identified
based on map-reduce and other computational algorithms which run on distributed
systems analyzing billions of entries in the resource logs. These identified
resources are stored/cached in server hash maps and the page is rendered in
phase wise manner.
This method and system is described in US patent 8,108,377
Predictive resource identification and
phased delivery of structured documents (Inventors: Jiang; Changhao, Wei; Xiaoliang; Assignee: Facebook,
Inc. (Palo Alto, CA)) .
As the disclosure goes on to describe Hadoop and Hive usage, we
find that
“…the resource logging,
analyzing, filtering, predicting, and/or selecting operations discussed above
can be implemented using Hive to accomplish ad hoc querying, summarization and
data analysis, as well as using as incorporating statistical modules by embedding
mapper and reducer scripts, such as Python or Perl scripts that implement a
statistical algorithm. Other development platforms that can leverage Hadoop or
other Map-Reduce execution engines can be used as well…”
Must say, this is one of the smart implementations of
predictive computation which is reducing latency, limiting network
load and overall leading to a better user experience. Remember, since this is
patented, check with assignee before any commercial usage.
Correct
ReplyDeletelink to patent updated...