Sepaton boosts disk-based backup speed

By Dave Simpson

Sepaton today released the 64-bit 6.0 version of its software and two new members in its S2100-ES2 series of disk-based backup systems with global data deduplication. The models 1910 (1TB SATA drives) and 2910 (2TB SATA drives) will compete primarily with high-end systems from IBM (ProtecTier) and EMC’s Data Domain unit (DD 890 and Global Deduplication Array, or GDA), as well as target deduplication systems from vendors such as Exagrid and NEC.

Sepaton hopes to differentiate the S2100-ES2 1910/2910 in part on performance, and has doubled throughput vs. predecessor models. Sepaton claims throughput of 1,500MBps per node, and a backup ingest rate of up to 43.2TB per hour on an eight-node configuration with 10GbE connectivity and Symantec’s NetBackup OpenStorage, or OST. (Sepaton’s backup systems previously supported OST, but not on 10GbE.)

“Eight nodes is the qualified maximum configuration today, but we will be qualifying more nodes – up to 16 – as users require it,” says Linda Mentzer, Sepaton’s vice president of product and program management.

For performance comparison, EMC cites 14.7TB per hour on its DD 890 (with OST Boost) and 26.3TB per hour on the DD GDA (with OST Boost).

Enterprise Strategy Group analyst Lauren Whitehouse advises users to closely factor in cost when comparing single-, dual- and multi-node target deduplication systems.

“You might think that a multi-node solution costs more than a single-node solution, but that’s not necessarily the case,” says Whitehouse. “Target deduplication systems are typically licensed based on capacity, so you can figure out a $/GB calculation to do an apples-to-apples comparison. With grid-based solutions users typically have a choice to license the solution based on performance and capacity, so then you need a $/MB/sec calculation to do an apples-to-apples comparison.”

Whitehouse also notes that multi-node grid architectures, such as Sepaton’s, can have potential advantages.

“Modular multi-node architectures allow users to start with what they need today and seamlessly grow,” says Whitehouse. “There’s no need to over-buy to future-proof the solution or under-buy if budgets are tight and companies are forced to only buy what they need today.”

In addition to the 64-bit operating system and throughput-enhancing OST, the performance boost in Sepaton’s S2100-ES2 1910/2910 is attributable to an upgrade to Intel’s six-core Westmere processors. Each 2U node includes two processors, four 4Gbps Fibre Channel ports, two 10GbE ports, and 48GB of RAM. The back-end architecture is an 8Gbps Fibre Channel SAN. Capacity scales to 1.6PB.

The S2100-ES2 1910/2910 platforms are based on Hitachi Data Systems (HDS) AMS 2100 disk arrays, which Sepaton adopted in mid-2010.

Other features include support for multi-tenancy in cloud environments, capacity quotas on a per-pool basis, and data compression in conjunction with global deduplication. In addition, Sepaton’s ContentAware technology deduplicates both progressive incremental Tivoli Storage Manager (TSM) backups and multi-streamed database backups.

Pricing for the S2100-ES2 1910/2910 starts at $257,500 for a single-node configuration with 12TB of capacity, four 4Gbps Fibre Channel ports and two 10GbE ports.

Related article:

Data Domain gets new boss array (The Register)

This article was originally published on January 24, 2011