IBM's new storage architecure to beef up data analytics speed
IBM unveiled its new storage architecture that doubles the analytics processing speed specifically suited for cloud computing and data intensive workloads.
The announcement came at the Supercomputing 2010 show. The new architecture is called the General Parallel File System-Shared Nothing Cluster (GPFS-SNC). The new architecture was developed at IBM Research-Almaden. It provides continuous operability for a long period of time through clustering technologies, dynamic file system management and advanced data replication techniques that keep data consistent across various users.
The new architecture is primarily suited for digital media, business intelligence, surveillance videos and financial analytics.
GPFS is the core of IBM's HPC systems, Scale-Out NAS (SONAS) and Smart Business Compute. The GPFS-SNC is a distributed shared-nothing architecture in which each node is independent and work is allocated to each node and no single node depends on other to start processing or forwarding.
It also uses Hadoop Distributed File System which evolved from Google's MapReduce framework that allows processing of large data sets in a distributed computing environment. However, Hadoop is a Java-based programming framework and thus cannot interact with Portable Operating System Interface (POSIX) based traditional applications, as the requirements of POSIX file system differs from the target goals for a Hadoop application. It is primarily used for applications involving search engines and advertising POSIX is based on UNIX OS and was developed to create standardization because enterprises wanted to able to develop programs that could be run on systems devised by other manufacturers.
IBM's GPFS is POSIX compatible and thus users with traditional applications that use POSIX interface can also use IBM's GPSFS-SNC technology in the cloud.
IBM stated that the new architecture can handle petabytes of data efficiently as the design provides a common file system and namespace across disparate computing platforms, streamlining the process and reducing disk space.
© Copyright IBTimes 2024. All rights reserved.