| |||
|
|
||
|
|||
| |
|
Home > Storage Technology News > Quantum leaps into file virtualization with StorNext |
| |
|
|
|
|
| Storage Technology News: |
|
||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
"At the core, it's still a SAN file system, but we're stretching it beyond that," said Nathan Moffitt, software product marketing manager for Quantum. "We're extending this product beyond just those [high-performance computing (HPC)] vertical markets and more and more into the general IT market." StorNext will now perform subfile-level data deduplication as it migrates files to archival storage, thanks to IP from Rocksoft, which ADIC acquired for $63 million shortly after the announcement that ADIC was to be acquired by Quantum in May 2006. Rocksoft's deduplication was also integrated with Quantum's DXI disk-based backup product in December.
Next, Quantum has added the ability to extend the file system to LAN clients via gateways that allows traditional network attached storage (NAS) devices running NFS or CIFS to have access to the proprietary StorNext file system through clustered gateways as if it were a local disk. This also means that LAN clients can connect to the SAN file system over Ethernet rather than requiring Fibre Channel and HBAs -- essentially vaulting StorNext into the world of file virtualization, according to Moffitt. Lastly, StorNext 3.0 now gives users the ability to run service operations or swap out storage devices while the system is running through the Dynamic Resource Allocation, new to this version of the software. Analysts said that even though this product has been around for years, Quantum is smart to reposition it now that shared file systems found in file virtualization and clustered NAS are gaining popularity in the enterprise. "I joked with Quantum that this product is one of the best kept secrets in the industry," said Dianne McAdam, analyst with the Clipper Group. "They said, 'yeah, we're trying to change that.' " StorNext is now competing with a multitude of different products, including Silicon Graphics Inc.'s (SGI) CXFS and DMF; IBM's GPFS and Tivoli Storage Manager (TSM); Cisco System's file virtualization through its recently acquired subsidiary NeoPath Networks; EMC's file virtualization through Rainfinity Inc.; Acopia Networks' and Attune Networks Inc.'s file virtualization products; clustered NAS, like that offered by Isilon Systems Inc., Panasas Inc. and PolyServe Inc.; and other SAN file systems, like Ibrix Inc., as well as file migration products from Moonwalk Inc. and Nirvana Storage. While most of its features stack up against competitors, it lacks at least a few features from each category. For example, Brocade Communications Systems Inc. and Cisco are positioning their file virtualization products alongside WAFS products for offsite replication, which StorNext also lacks. As a tiered storage product, it also lacks specialized archival features, such as WORM and data classification features, and searches for legal discovery. "This is a very difficult product to classify," McAdam said. "You can't put it in one bucket." Unlike parallel file systems, like Isilon's SyncIQ and Panasas's ActiveStor, StorNext handles all traffic through the metadata controller, which users must assemble with separately bought hardware, though Quantum said it will also work with users to determine the right server to support their environment's traffic requirements. According to analysts, users should be careful when it comes to getting the correct hardware for the metadata controller, which otherwise could turn into a bottleneck. "Users need to be careful on their hardware selection," McAdam said. But, she pointed out, "[StorNext] has some track record on performance -- customers have been pumping a lot of data through [the metadata controller] in high-performance computing environments in fairly large accounts." According to Brian Garrett, technical director of the Enterprise Strategy Group Lab, his own tests of the product showed a minimal performance impact from deduplication. However, he warned, "files in production are not as deduplicable as backup, where you tend to see a lot of the same data over and over. This is why Quantum is claiming a more modest 10-to-1 savings than the backup deduplication vendors who are claiming 20 and 30 to 1." Garrett said ESG Lab tested the deduplication of data using MPEG movies taken from seven days' worth of data from a weather surveillance camera in Arizona. The test began with around 8 GB of MPEG data, and Garrett said that data was reduced 40% by StorNext's deduplication. "I was impressed with that figure," Garrett said, "Because MPEG data is already compressed." Garrett said the addition of the LAN client extension boosted performance over NAS-attached servers running NFS in his lab from 73 megabytes per second (MBps) to 111 MBps using one Ethernet line and up to 210 MBps with dual Ethernet connections. ESG Lab also tested out the ability to swap out storage devices while the file system was running. Garret said his team watched a movie and copied files on one area of the file system while swapping out the underlying storage. He said there was no disruption during the test. However, he said, "I was surprised to find out they didn't do this before -- maybe it wasn't high on the list of priorities for research computing." Garrett also pointed to a potential flaw in Quantum's repackaging of the product for the enterprise market in general: assembly is still very much required. "Users need to work with the vendor or systems integrators to bring together all the right pieces," Garrett said. "There's no place a user could go and say, 'I need 15 terabytes capacity in the file system with the storage to match' and be able to buy the whole thing in a rack. It would be nice to see them offer a variation on the product like that."
|
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||
|
|||||||||||||||||||||||||||
| |||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||
