|By Unitiv Blog||
|December 28, 2012 03:30 PM EST||
Long a leader in storage technology, IBM last summer gave us a peek at some of the advances they’re making in storage. Today, IBM has created a high-speed storage system than can scan over 10 billion files in just under 45 minutes. The previous system from IBM scanned just 1 billion files over three hours.
At the heart of this increased performance was the use of flash memory. The flash memory is used to store metadata in the storage system, and that meta data is used to find requested information. Usually, metadata lives on the disk, which can greatly slow down storage operations.
A coming need
IBM recognize that their customers will be dealing with more and more data over the next few years. As customers need to handle and store larger and larger amounts of data over longer periods, they need even more efficient ways to process and handle that data.
The demonstration last summer used a cluster of 10 servers, each with eight cores. The cluster had access to a total of 6.8 terabytes of flash solid state memory, provided by four Violin Memory Storage Systems devices.
IBM’s own GPFS (General Parallel File System), which was developed originally to be used in high-performance environments, was used for this system. GPFS is being today used for more than just high-performance environments, and is finding use in data-intensive workloads in the enterprise. The system lets processor cores to be able to write and read from disks in a parallel fashion, which greatly increases the responsiveness of the storage system.
Where it will go from here
It’s likely that, in the coming months, we may see some implementations of this research in IBM’s own product lines. Solid state systems have led in the past to the creation of IBM Easy Tier, a storage management system that lets users balance their data loads between both solid state disks and regular disks.
Accordingly, other companies are showing interest in how to use solid state to speed up their operations. A recent study from Carnegie Mellon and Intel Labs proposed a particular server architecture that would pair low-power processors alongside flash memory, potentially speeding up websites that are transaction-heavy.