Using Hadoop for data on Google's cloud? Google would rather you didn't
And it's got just the replacement for it: a shiny 'Google Cloud Storage Service'
Google wants to shift heavy users of its cloud services away from an open-source, community-developed filesystem and into its own proprietary Colossus tech.
The upgrade was announced by the web overlord in a blog post on Tuesday that announced admins could now store Hadoop-destined data directly in Google's closed-sourced Colossus-based "Google Cloud Storage Service", and threw mud at the traditional Hadoop File System (HDFS) plugin.
The service, we're told, provides a more efficient connector between Google's cloud storage and compute services, and represents another advance in the Chocolate Factory's rent-a-server infrastructure which competes with Amazon Web Services and Windows Azure.
Hadoop is an open-source data analysis platform based on ideas outlined in the Google File System and Map Reduce papers which came out of Google in the early 2000s.
Since Hadoop's genesis at Yahoo! in the 2000s it has become a standard component of any data analyst's open-source toolkit, and its development is stewarded by companies including Cloudera and Hortonworks.
Google, though, would prefer it if users of its cloud opted for the closed-source Colossus-based Google Cloud Storage. To tempt them over to the system, it has listed some of the benefits of using Colossus over HDFS. These benefits, according to Google, include "no storage management overhead", "high data availability", and "quick start up."
Colossus has multiple master nodes which gets around some of the redundancy problems that bedevil early HDFS implementations. It also uses Reed-Solomon erasure codes to perform error correction which, Google says, "achieve similar resilience to failures compared to replication, though with less storage overhead."
Developers should bear in mind that using the cloud storage service locks them further into Google's own idiosyncratic way of doing things and pushes them further away from the main filesystem of the open-source large-scale data community. ®