By Kevin Komiega
When Network Appliance acquired Spinnaker Networks in late 2003 it had plans to merge its own Data ONTAP operating platform with the global namespace and clustering capabilities of Spinnaker’s SpinFS software. More than two years later the acquisition has borne fruit with the release of a new storage platform for the high performance computing (HPC) market called Data ONTAP GX.
NetApp last month began shipping the Data ONTAP GX operating system with the hopes of recruiting new customers that require massive performance, capacity, and scalability. NetApp officials say the new platform is ideal for customers using applications for seismic processing, digital content creation, visual effects rendering, and chip design and simulation.
The single global namespace capability of ONTAP GX, which came from Spinnaker, enables multiple storage nodes to be presented to applications as a single system and data to be moved between storage nodes and tiers transparently, which simplifies data retrieval and management tasks, according to Rich Clifton, vice president and general manager of NetApp’s networked storage business unit.
Data ONTAP GX works with Net-App’s hardware platforms, including the recently introduced FAS6070 and FAS3050 systems. HPC customers can use the clustered file system technology in Data ONTAP GX to stripe individual files or datasets across multiple nodes for better performance.
NetApp’s Clifton says the new version of the operating system is aimed at the HPC market for several reasons. “People are aggregating very large amounts of compute capabilities to solve large computational problems and these environments are very data-intensive,” he says.
The computing and capacity requirements of HPC customers are nothing new. There has always been an expectation among these types of customers for “bleeding-edge” performance and scalability. But Clifton says the scalability of Data ONTAP GX-the first release scales to about 6 petabytes under a single global namespace-and its ability to utilize virtualization can help HPC customers change how they predict and resolve problems.
NetApp claims performance of more than 1 million operations per second, based on the SPEC SFS benchmark. Data ONTAP GX scales in a modular way with the option of striping volumes for massive throughput on a single file.
“ONTAP GX has the ability to expand a storage environment by incrementally adding disk storage and processing power to the underlying environment, and it becomes part of the virtualized pool of capacity and performance,” says Clifton.
NetApp combined pieces of its own operating platform and elements of Spinnaker’s SpinFS. For example, Data ONTAP GX features the snapshot, volume management, and data-protection capabilities of the ONTAP platform alongside the global namespace and cluster scale-out technology from Spinnaker. The NetApp WAFL file system is at the heart of Data ONTAP GX, and NetApp’s RAID-DP provides tolerance to multiple simultaneous disk failures.
Microsoft ships HPC OS
High performance computing (HPC) technology might not just be for bleeding-edge companies and scientific research firms. Microsoft officials say they are bringing HPC to the masses with the release of Windows Compute Cluster Server 2003, the company’s first offering designed to run parallel, high-performance applications.
Windows Compute Cluster Server 2003 is aimed at classic HPC customers in engineering, exploration, life sciences, scientific research, and other industries that need computational power and scalability to solve complex computations, simulations, and analyses. But Microsoft hopes to make HPC technology more mainstream by bringing the cost advantages, management features, and partner ecosystem of the Windows Server platform to a broader range of users in commercial industry and the public sector.
A number of Microsoft partners are scheduled to publicly release solutions that run on, or interoperate with, Windows Compute Cluster Server 2003 by year-end, including ABAQUS, Absoft, Advanced Micro Devices, ANSYS, BioTeam, Broadcom, CD-adapco, Cisco, Dell, ESI Group, Fluent, Fujitsu, Hitachi, Hewlett-Packard, IBM, Intel, Livermore Software Technology, Macrovision, MathWorks, Mecalog Group, Mellanox, MSC Software, Myricom, NEC, Parallel Geoscience, Platform Computing, the Portland Group, Schlumberger, SilverStorm Technologies, Tyan Computer, Verari Systems, Voltaire, and Wolfram Research.
Windows Compute Cluster Server 2003 is currently shipping to OEMs with volume channel availability scheduled for next month. The software is priced at about $469 per node. Evaluation versions of the software are available now on Microsoft’s Website.