Cloud Computing: EMC Touts Isilon for Hadoop
February 8, 2012
EMC moved to make Hadoop safe for the Joe Blow big company that wants to figure out what’s in all that unstructured data it’s got.
It’s integrated the Hadoop file system into its Isilon scale-out NAS storage and lets the two of them cuddle up with its Hadoop-digesting Greenplum analytics machine. Greenplum also handles structured data so users can run queries across both datasets.
EMC says that makes Isilon the very first enterprise-class scale-out NAS system with native Hadoop support, giving Hadoop Big Data enterprise comfortable features like snapshots, replication and backup…
The storage Goliath figures it can probably make a packet off the scheme, removing Hadoop’s NameNode single point of failure and making it manageable by mere mortals.
The way EMC does things Isilon’s native petabyte-scale OneFS file system actually handles the Big Data, not Hadoop, thanks to Isilon’s native OneFS file system being integrated with Hadoop’s Distributed File System protocol and chewing through the scads of resident in the scale-out repository.
HDFS ceases to be a file system and turns into a protocol so Hadoop can inherit the high availability characteristics of OneFS and comes free to users as a maintenance update to the OneFS 6.5 operating system.
EMC reflects that "early adopters of Hadoop have relied on makeshift storage infrastructures not optimized for Big Data exploration, inhibiting the shared data access critical to robust analysis, and preventing wide adoption of Hadoop in the enterprise" beyond Facebook, Amazon, Twitter and Adobe.
Isilon is supposed to remove the integration complexities of disparate open source components and hardware.
EMC’s version of Hadoop uses the MapR file system although EMC users can apparently use whatever Hadoop distribution they want.
EMC bought Isilon in July of 2010 and Greenplum the following November. It has yet to buy MapR.