StorNext Raises the Data Management Stakes

With the release of StorNext 3.0, Quantum makes the most of its ADIC acquisition. And de-dupe is just the start...
Posted August 9, 2007

Drew Robb

Drew Robb

Since its acquisition of ADIC in 2006, Quantum Corp. of San Jose, CA, appears to have done a better job than many other storage vendors in integrating the products of the acquired company into its portfolio. ADIC’s Scalar and StorNext products, in particular, are being marketed heavily since the deal and appear to be front and center in the company’s product development roadmap.

Most recently, Quantum released StorNext 3.0. This latest version extends data sharing to servers on the LAN and integrates data de-duplication into the software.

Three Areas of Upgrade

StorNext has been upgraded in three core areas. Data Reduction Storage (DRS) is a specialized disk tier that incorporates de-duplication. DRS tunes data reduction to a specific data set and increases the likelihood that very high levels of data reduction can be achieved. DRS is established on a local StorNext volume attached to the MetaData Controller (MDC), which serves as the traffic cop – handling disk allocation as well as client side buffering so that when multiple clients are reading or writing to/from the same file, they all see the same content. This architecture guarantees high throughput at FC speed rather than LAN speed.

“Data De-duplication reduces a customer’s data footprint, saves money by lowering capacity requirements and enables data to be retained on fast recovery disk for a much longer period of time,” says Nathan Moffitt, StorNext product manager at Quantum. “Data reduction rates of 10x or more can be achieved, depending on the data type and amount of data in the de-dupe volume.”

StorNext’s Storage Manager includes an integrated policy engine that automatically moves files from primary disk to one or more storage repositories (including tape). Thus data movement policies can also be established to manage the migration of files from primary storage to DRS and back. Further, primary disk can be freed up easily for high priority tasks.

Another addition to StorNext 3.0 is Dynamic Resource Allocation (DRA), which increases uptime by enabling online service operations. It allows customers to scale their storage – adding new storage capacity or transparently swapping out disk arrays during hardware upgrades – while the system is active.

DRA also enables rapid data sharing among servers by virtualizing the underlying storage components. A function known as File System Expansion allows capacity upgrades without disrupting business operations. Disk capacity can be added to an existing StorNext file system on the fly. The customer defines new LUNs and makes them available across the SAN to the MDC and StorNext clients. All I/O currently in progress remains unaffected during the expansion.

Another feature, Move Stripe Group Data, permits data to be moved from one disk volume to another. If, for instance, an array doesn’t have enough throughput or capacity, data can be quickly shifted onto another array.

This function can also be utilized to smoothly retire older arrays.

“StorNext can automatically move data from one stripe group to other stripe groups,” says Moffitt. “During migration, the file system is left online and read/write operations occur normally.”

Distributed LAN Client (DLC) is the third major area of StorNext upgrade. It enables applications on the LAN to access a shared pool of storage faster, and with a higher level of resiliency, than most traditional network sharing methods can provide. Using clustered gateways for access and an optimized communication protocol for performance, LAN Client extends data sharing, and delivers load balancing and transparent I/O failover.

“StorNext Distributed LAN Client allows LAN-based servers to connect to via clustered gateway systems,” says Moffitt. “To maintain the highest level of throughput, they only service LAN I/O and do not run applications.”

The LAN Client communicates with the MDC to determine file location, disk allocation and buffering. Once the LAN Client knows a file’s location, it passes on data request. The gateway system retrieves the data and passes it to the distributed LAN Client.

While similar to CIFS / NFS sharing used in NAS, DLC differs by utilizing a specialized TCP/IP protocol designed for higher per-stream throughput. Single GigE connections using LAN Client have been observed at 110 MBps, a marked improvement over NFS or CIFS.

Additionally, this protocol balances I/O across available gateways to avoid “hot spots” which could bottleneck throughput. It can also fail over I/O between gateways if one is unavailable.

All this adds up to an improved storage management platform for Quantum's customers.

“Thousands of existing customers in the rich media and HPC markets rely on StorNext software for fast access to shared file systems, improved workflow and automated HSM,” says Brian Garrett, an analyst with the Enterprise Strategy Group of Milford, MA. “The new and improved StorNext will also be appreciated by enterprises looking for a better way to share and manage unstructured digital assets.”

This article was first published on

Comment and Contribute


(Maximum characters: 1200). You have characters left.