Seagate heading for the wild west
A romp with cowboy Bill
Seagate flash drives
Seagate wants to have its own solid state drive (SSD) intellectual property (IP) above the raw componentry. It doesn't want to own a foundry. The flash fabs don't make money and, anyway, NAND flash technology suffers from short write cycle life, poor write performance relative to reads and high prices. Single level cell (SLC) technology is the best performer but is expansive and can't scale. Multi-level cell (MLC) technology has less expensive capacity but suffers particularly from the reliability issue.
Seagate's coming SSD will combine SLC and MLC flash in the same unit. Watkins said: "We want to put performance in the read-mode, tier zero, enterprise space. That makes business sense to us initially."
The drive will be for tier zero applications needing lots of low latency IOPS, the applications that use short-stroked 15K rpm Fibre Channel drives today. Compared to the cost/GB of such drives an enterprise flash drive is much better value. Seagate reckons this market could grow to 5-10 per cent of enterprise storage purchases.
It will continue to prosecute its legal IP infringement case against STEC and says cross-licensing isn't feasible. STEC doesn't have any IP it needs, Seagate already owning Fibre Channel and SAS interface and error detection and correction technology.
It thinks that the big prize in solid state is next-generation technology like magneto-resistive memory (MRAM) or spin transfer torque RAM or similar that will surpass NAND limitations. NAND will give Seagate the base from which to grow into this next-generation technology. There are a number of Silicon Valley start-ups in this area into which Seagate has invested or with which it has joint-ventures. A purchase of SanDisk for this kind of technology looks unlikely.
Hybrid drives, areal density growth, etc.
Seagate will introduce hybrid drives, hard disk drives (HDD) with, say, 15GB of flash, and the ability to boot from flash to solve the long boot time problem. Such drives would also cache writes in the flash and so reduce the number of HDD writes, lowering power consumption. This could happen next year and would, presumably, be particularly good for notebooks.
Watkins said he expects areal density to improve at 50 per cent plus per year. A 1Tbit/sq inch level should be reachable with product appearing in four to five years (2012-2013). In 2013 heat-assisted magnetic recording (HARM) technology will start being used.
Seagate has looked at 20K, and even 25K rpm drives but the cost model has never seemed to be right with the added expense for the added performance being too high except possibly for a very small niche. Anyway, SSDs have taken the pressure off doing anything here for now.
Yes, there is scope for more consolidation in the HDD industry. Of the six main players only Seagate, Western Digital and Hitachi, all vertically integrated, are making money, with Fujitsu, Samsung and Toshiba losing money and being "extremely unhealthy". None of this sick trio are vertically integrated with Watkins saying the vertically-integrated model is the only way to scale to any size. Overall Seagate thought it was well-positioned to take advantage of long-term storage trends, almost being recession proof. The fourth quarter of this year will see a blip with growth lower than last but pretty soon it should be business back to normal.
And the answer to the most pressing question? Bill voted for Obama. ®
Sponsored: Hyper-scale data management