October 5, 2008 Leave a comment
Interesting that they were using only 1.6 % of the 14.25 PB capacity of the HDFS Cluster. Even more interesting is that they already have 226 dead nodes (and 4049 live nodes,) which is a 5% attrition rate so far, whatever ‘far’ means in this case, I would be interested to know.
In other Hadoop news, the “Hadoop + Python = Happy” project looks interesting. Basically it is a framework for writing map-reduce programs for Hadoop using Jython (basically Python.)