Database gurus slammed for Google post
MapReduce: a "major step backwards"
A database pioneer and honored computer science professor have come under heavy fire for issuing a strong critique of Google's MapReduce technology for processing large unstructured databases.
Ingres inventor and Postgres architect Mike Stonebraker and his colleague, University of Wisconsin computer science professor David DeWitt, have been accused of "not getting" data in the clouds while others have demanded the duo retract what's been branded a "highly inaccurate article".
They called MapReduce a major step backwards because it is "sub optimal", lacks the features commonly associated with database management systems (DBMS) and is incompatible with "all of the tools DBMS users have come to depend on". They also said that it is not '"novel". They conclude that MapReduce ignores many of the developments in parallel DBMS technology over the last 25 years.
Their joint blog post drew fire from bloggers and a barrage of commentators coming out in support of MapReduce, including a detailed riposte that claimed DeWitt and Stonebraker don't know what they are talking about.
The gist of the counter argument is that MapReduce can't be compared to a relational DBMS because it is a technique for dealing with large amounts of unstructured data rather than the formal tabular data in relational DBMS. Google reckons it processes 20PB of unstructured data a day using MapReduce.
The almost complete lack of support for the view put forward by DeWitt and Stonebraker suggests they might well have misunderstood MapReduce's role in modern data processing.
Given the eminent background of both academics, though, this is surprising. DeWitt has researched large parallel DBMS since the 1980s and, in addition to his pioneering work on Ingres and Postgres, Stonebraker is currently active in the large DBMS area with his new company Vertica.
DeWitt has published more than 100 technical papers and been honored for contributions to database systems having started in the mid 1970s on a NASA- and DARPA-funded project looking at scalable object-relational system for managing very large geo-spatial data sets.
Requests for an interview have yet to be answered.®