For researchers plowing through some of the toughest scientific and large-scale computing challenges, system performance is only part of the problem. With the influx of big, diverse datasets, the role of high performance storage is becoming increasingly important.
NetApp E5500 storage system |
To address these issues, NetApp and SGI made simultaneous announcements yesterday of new storage systems specifically targeted at HPC and big data applications. NetApp launched its latest offering in the E-Series line of controllers yesterday, the NetApp E5500. The E-Series is the product line from Engenio, which NetApp bought from LSI in early 2011.
SGI, an OEM partner of NetApp, has been incorporating the E-Series products into its RAID systems for years. In fact, the company has already adopted NetApp’s new E5500 in its InfiniteStorage 5600 RAID system, also released yesterday, and it has made at least one pre-sale to an Australian cosmological research center.
NetApp says the E5500 has the highest speed, scale, cost-efficiency and scalability on the market, along with a very small footprint. To back up these claims, SGI showed off some audited, peer-reviewed SPC-2 benchmark results for its RAID systems using the controller. As for speed, the IS5600 achieved a per-spindle throughput of over 70 MB/second, some 2.5 times higher than the second-fastest product (from Sun).
“It’s all about performance for this set of customers,” says Bill Peterson, senior manager for Hadoop and big data analytics at NetApp. “Our customers have traditionally built all their systems around highly structured database schemas. Then you go to something like Hadoop, for example, and there is no schema. They want performance. At the end of the day, whether it’s for big data or HPC, they’re trying to gain insight, and that insight is always required quickly.”
To reach its target audience, NetApp also offers preconfigured solutions optimized for applications such as high performance Lustre, full-motion video, media content management and seismic processing. With the IS5600, SGI also offers dedicated, application-specific support for parallel file systems, including Lustre, Hadoop and CXFS.
The E5500 “reaffirms NetApp’s commitment to high-performance storage, which is important for organizations seeking to balance HPC and big data scalability and performance with the reliability of an enterprise-class storage provider,” says Addison Snell, CEO of Intersect360 Research. “This release also underscores the ongoing maturation of the Lustre community, which is essential to bringing parallel file system technology to a wider range of users.”
Speed isn’t everything. SGI also shows off SPC-2 benchmark tests that give its RAID product the best cost-efficiency on the market, at about $15 per MBps performance. That’s about half the cost per MBps of the next most efficient system, from Fujitsu.
SGI is also targeting its IS5600 RAID platform at HPC systems. Both the IS5600 and its previous generation IS5500 (based on NetApp’s E5400 storage system) can scale to 384 drives and 1.2 petabytes, but the IS5600 offers 2.5 times more I/O. The combination of NetApp’s storage technology and SGI’s RAID configuration “really blazes a new trail, especially for price/performance,” says Floyd Christofferson, director of storage product marketing at SGI. “You can’t beat it for HPC.”
Christofferson is also impressed with NetApp’s SANtricity Dynamic Disk Pools, or DDE, which provides improved fault-tolerance for RAID. DDE uses a proprietary algorithm to distribute data, parity information and spare capacity across a pool of drives. Unlike RAID 6, which can generally only tolerate one drive failure before having to take the RAID offline to replace faulty drives, with DDE many drives can fail and the system can recover very quickly without being taken offline.
“You can keep failing disks all day long” says Christofferson. “As long as you have more than 11 drives in there, you’re good to go. It will come back to performance within minutes, as opposed to hours or days” for RAID 6. “This could be a viable successor to RAID 6,” he says.
SGI won’t be the only company debuting systems with NetApp’s new devices. IBM, RAID Inc., Teradata and others are all NetApp OEM partners.
But Christofferson says SGI already has “several customers in the funnel,” for its new IS5600, which is expected to start shipping this April. One firm order is for a research facility in Australia, conducting very high-end research on fields such as general relativity; galaxies, cosmology, dark matter and energy; and even “transient Radio Phenomena caused by Extraterrestrial Life.” That facility will be deploying a 100PB system from SGI this summer. It will use 27 IS5600 systems with 6 PB of storage and 94 PB of virtual storage.
It takes a lot of fast storage to read messages from ET calling long distance.