A partnership between Intel and Hitachi Global Storage Technologies has missed the ship date for its first enterprise flash storage products, but that stumble wasn't a harbinger of bad times for the technology, which continues to gain acceptance in data centres, an industry analyst said Monday.
The companies announced a deal in late 2008 to jointly develop SSDs (solid-state drives) for enterprise servers and workstations. Hitachi plans to use flash silicon developed by the partnership of Intel and Micron in drives with SAS, or Serial Attached SCSI, and Fibre Channel interfaces. Less than a year ago, Hitachi GST Director of Enterprise Marketing Dean Amini said the companies were on track to deliver products in the first half of 2010.
Initial products now are scheduled to come out in the second half of this year, said Ulrich Hansen, director of market development at Hitachi GST.
"We did have a slightly more aggressive target initially set, but we're still very enthusiastic about this partnership," Hansen said in an interview on Monday. Hitachi is a major OEM to server and workstation manufacturers. These will be its first SSD offerings.
Hansen spoke on the eve of the Flash Memory Summit conference in Santa Clara, California, where executives and customers will gather to discuss the range of flash technologies from consumer electronics to data-centre storage arrays.
Per gigabyte, flash storage is more expensive to make than HDDs (hard disk drives), but it typically uses less energy and allows for quicker retrieval of data because it has no moving parts. Enterprises that have adopted SSDs generally have used them for data that needs faster retrieval, while they left to HDDs larger amounts of content that isn't as urgently.
The enterprise flash business as a whole, which has faced some stumbling blocks over the years, is still growing fast, according to analyst Jim Handy of Objective Analysis.
"The market is really budding right now," Handy said. His company forecasts global SSD shipments to rise from 170,000 this year to 4.1 million in 2015 while revenue increases from $425m to $3.8bn.
SSDs long suffered a reputation as not reliable or long-lived enough for enterprise use, but much of the problem had to do with the controllers that were used to manage the flash in earlier products, Handy said. Some used controllers that were designed for consumer Compact Flash cards, which get much less intensive reading and writing use than enterprise storage does, Handy said. With purpose-built controllers, SSDs don't represent a step down from the reliability and long life of HDDs (hard disk drives), he said.
SSDs are likely to remain more expensive than HDDs for the foreseeable future, but enterprises will pay the premium for their speed where it is needed, Handy believes.
Intel and Hitachi fell behind schedule because some development took longer than expected and they have added more features and carried out more testing of the products, Hitachi's Hansen said.
"We are essentially finished with the development work. We're validating the product [and] we're doing pre-qualificaiton work with some of our OEM customers," he said. The first of the products from those OEMs are likely to hit the market before the end of this year, according to Hansen.
As planned, the products will use 6G bit-per-second (bps) SAS and 4Gbps Fibre Channel interfaces. The first SSDs from the partners will use SLC (single-level cell) flash, the form that has been used in most enterprise SSDs so far. SLC packs less data into the same amount of space and is more expensive per gigabyte, but it also has greater longevity for uses where data is intensively written to the drive, Hansen said. Ultimately, it is less expensive for these applications, he said. By the second half of next year, Hitachi and Intel expect to come out with versions that use MLC (multilevel cell), which is more dense, less expensive and adequate for applications where data is written less intensively, he said.
Consistent with Amini's forecast last year, the drives will have an MTBF (mean time between failures) of as much as two million hours, Hansen said.