IBM has unveiled two new enterprise-class storage arrays which, despite being aimed at different markets, can be managed through a single interface and data can be replicated between them.
The TotalStorage DS8000 is IBM's next generation of its high-end Enterprise Storage Server, also known as Shark, but with up to six times the performance and more than three times the capacity at 192Tbytes. The refrigerator-size array also comes with 256Gbytes of cache memory and 128 Fibre Channel or Ficon ports.
The TotalStorage DS6000 array is a rack-mountable, 3U-high unit that scales from 580Gbytes to 67.2Tbytes. It comes with a 4Gbyte cache and 16 Fibre Channel or Ficon ports. Starting at $97,000 (£54,000), the DS6000 is IBM's lowest-cost midrange product to date. A two-processor, 250Tbyte model of the DS8000 starts at $134,000 retail.
Both boxes can be used as primary storage for IBM zSeries and iSeries mainframe computers as well as Unix, Linux and Wintel servers. The DS8000 can be upgraded in the field from two to four processors.
For the first time, IBM is also using its server processors for storage, which cut its time to market in half. While the DS8000 uses the P-Series processor, the DS6000 uses an IBM Power PC processor.
IBM also increased the industry-standard warranty on the arrays from three years to four years. IBM said that over the next year, it expects to make the DS8000 expandable to eight processors and 512Gbytes of cache, and will also offer a mix of Fibre Channel and cheaper Advanced Technology Attachment discs.
Tony Asaro, an analyst at Enterprise Strategy Group, gave IBM high marks for the new disc arrays for compatibility, the four-year warranty and the use of high-end IBM server and PC microprocessors.
"This is a great product family for the storage-area network," Asaro said.
But he added that IBM still needs to address the network-attached storage and content-addressed storage markets, as have its two chief competitors, EMC and Hitachi Data Systems.
The DS8000 has logical partitioning, which allows users to split the box into two systems that can run completely different operations. Both arrays use an internally switched architecture instead of Fibre Channel arbitrated loop, which brings IBM up to speed with similar boxes from competitors such as Hitachi and EMC.
Lucas Mearian writes for Computerworld