Comment Disks break - everyone knows that. Yet storage array vendors have rejected a technology that would get around this and save their customers pots of money.
This resulted in Seagate writing off millions of dollars, and customers continuing to spend money buying products from storage vendors that could have been much more reliable.
Spinning disks with read/write heads darting in and out across the platter surfaces suffer mechanical vibration and break. The data on them becomes lost and has to be regenerated through RAID schemes with the failed drive being replaced. Sealed canisters of drives allow individual drives to fail in place, with the canister controller logic recovering from the drive failure such that no data is lost and no engineer service visit is required.
Xiotech, one of only two companies (the other being Atrato) that provide sealed canister arrays, offers a remarkable five-year warranty on its Emprise product using its ISE (Intelligent Storage Element) technology, something no other mainstream storage vendor does.
Seagate, HP, IBM and start-up Sherwood Information Systems (which became Atrato) have each looked at the sealed storage canister idea. What happened? The story of how the technology was invented, developed and then rejected, by suppliers who say they listen to their customers, illustrates how business models reject disruptive technology.
Six years ago IBM Almaden was developing its IceCube technology, a three-dimensional storage brick, the IceCube or Collective Intelligent Brick, containing disk drives and intelligence. Each brick contains extra capacity so that disks can fail in place and no one need come and fix the system.
IBM thought that a brick could run for four years with a three per cent component failure rate. Bricks communicated to other bricks through cable-less capacitative coupling, and such technology would mean very much lower service costs. But here in 2009, no IceCube products have been announced by IBM.
In 2007 IBM was reportedly spinning off the IceCube technology into Seval Systems, a separate company. It did not want, it seems, to use the technology iteself. Seval Systems has since disappeared, as has the technology. Eric Wendel, the founder of Atrato, then Sherwood Information Systems, reckons there were "untold millions spent in development but never monetised" by IBM with IceCube.
In 2006 HP had its Federated Array of Bricks (FAB) project. These bricks would be composed of disk drives, a controller and communications cards and would federate themselves to provide storage volumes to users. A FAB product would offer the reliability and performance of enterprise-class disk arrays, at a fraction of the cost and with better scalability. Yet there are no HP storage products using this technology today.
So two existing mainstream storage array suppliers looked at sealed canister technology and rejected it. What happened with similar projects inside the world's biggest disk drive manufacturer and at the start-up?
Seagate's activities in the sealed canister area began with an ACORNS project before 2003. That year it started up its Advanced Storage Architecture group by recruiting a bunch of people from Compaq, including the then chief technology officer (CTO) for StorageWorks, Steve Sicola, and Dr Ellen Lary, business-critical storage VP. Compaq had inherited these people when it bought DEC (Digital Equipment Corporation). In effect Seagate bought a Sicola-led technology development group, and it started work at its Colorado Springs base on the ISE technology.
That same year, Eric Wendel recruited Dan McCormick and Jonathan Hall, and they started up Sherwood Information Services in Minnesota to develop a SAID (Self Maintaining Array of Identical Disks) disk drive array technology product. This built on work Wendel had already done to find out how to pack commodity SATA 2.5-inch laptop drives close together and produce an inexpensive but very reliable storage array for customers needing to access and store massive amounts of data. The company was renamed Atrato in 2004.
Solving the rotational vibration problem
Packing disk drives closely together can expose each drive to vibrations from its neighbours and so shorten its life. The sealed canister technology developers had to solve this problem, which they attributed to the spinning platters inside the drive, to the drive's rotational vibration (RV).
Sicola's ASA team did lots of development work, and in April 2004 filed its first ISE patent. This design, Wendel believes, did not work because Seagate's people did not understand the source of the rotational vibration. Many, many more Seagate patents followed with one in July 2004, describing a design in which disk drives were fixed in oppositing orientations so as to cancel out RV.
It was not until June 2005 and patent number 11/145, 404, that Seagate designed a technology to cope with the true source of RV, the high-speed acceleration and deceleration of the drive's actuator assemblies during seek operations.
The patent text stated:
The actuator seek torque during acceleration and deceleration imparts rotational vibration stimuli to the enclosure which, in turn, creates an opposite reaction torque. The reaction torque can create rotational vibration interference that adversely affects the data storing and retrieving performance of the device itself, and of adjacent devices in the same support structure. Where a plurality of the devices are grouped together in (an array), the rotational vibration interferences are compounded.
A mass sink is used to deal with this. Wendell says Atrato - he himself - had this problem solved almost 18 months before. His December 2003 patent application, number 60/533,605, stated: "Position drives... such that seek-caused actuator rotational-acceleration vibration causes simultaneous canceling rotational torque."
With the RV problem covered, Seagate completed its ISE development and started trying to sell the technology concept to its OEMs from 2007 or so onwards. But not a single one, in America or Japan, wanted it.
Seagate's misunderstanding of its OEMs' needs
According to Wendel, Seagate's justification for the ISE project was that it was a defence against the commoditisation of Seagate's high-end, enterprise-class, disk business, then Fibre Channel drives but now including SAS drives. These disks would be containerised in a proprietary fashion, and so create a value proposition exclusive to Seagate.
The company postulated that its OEMs (EMC, NetApp, etc) would be willing to give up their array subsystems business, and wanted to do so because they saw their value lay in storage software. Seagate reckoned they'd be happy to use horizontal, scalable building blocks with integrated RAID and other functionality.
"But," Wendel says, "when you look under the covers at where Seagateâs enterprise disk OEMs really make their money (and even now, not one of them wants to admit it), their revenue and profit models were then and still are heavily dependent on selling and maintaining proprietary hardware â including disks and enclosures.
"So, in reality, the Seagate 'bricks' initiative was from the outset extremely disruptive to the 'mainstream market models for making money' among Seagateâs most important OEM customers."
Wendel thinks Seagate's OEM customers rejected the ISE technology because "it threatened to commoditize their entire subsystem value propositions and (probably worse) it dissipated the value of the hardware service and support side of the array vendors' business, a key recurring revenue stream.
"[Seagate's] failure to understand the extent to which ISE would disrupt the very customers they were targeting meant that the product was essentially doomed from the beginning."
Another source close to the situation at the time has talked of discussions, sometimes heated, between Seagate and its OEMs, and of one OEM suggesting it might buy its disk drives elsewhere if Seagate persisted with the ISE technology, but this has not been substantiated elsewhere.
Neither Seagate nor any of the storage array suppliers have corroborated this view of events. One array vendor source said that locking itself in to proprietary Seagate technology was not an attractive proposition when compared to their existing ability to dual or triple source disk drives.
Seagate's ISE exit
In the event Seagate decided not to proceed with the ISE project and managed to sell its Advanced Storage Architecture group to Xiotech in July 2008.
One problem; Xiotech didn't have enough money. Step forward Stephen Luczo, then Seagate's chairman, who himself leads a finance round that results, a couple of weeks later, in a $40m equity investment in Xiotech. Luczo becomes a member of Xiotech's board and Sicola's group continues its work at Colorado Springs under Xiotech ownership and management; Sicola becomes Xiotech's CTO and Lary runs ASA inside Xiotech as its Advanced Solutions Group.
"I think it's probably fair then to say then that ISE is already among Seagateâs biggest product failures," says Wendel. "And thereâs a common theme here. IBM's fantastic "IceCube" technology (and) Seagate's ISE - huge development costs and five years of high-level strategic focus from the biggest storage company on the planet, yet it never generated any revenue for Seagate and was ultimately sold off (presumably at a big loss).
"Both of these products had the potential for delivering a huge step-function increase in data-center reliability and managability, but even with IBM and Seagate behind them, they failed. End-user customers want product reliability, but letâs face it â fixing stuff that breaks reliably is a profitable, recurring revenue stream."
How have they done since then?
In March 2008 Atrato, now located in Westminster, Colorado, launched its Velocity 1000 array. Xiotech launched its Emprise array product in April that same year. Two companies launching what looked like roughly similar products but with independent IP.
Wendell himself left Atrato before the product was launched and is now working at a stealthy start-up developing technology he invented while at Sherwood Information Systems. Subsequently Atrato has announced sales to about 15 customers, not a lot, and sought more investment funding.
Xiotech has enjoyed much more sales success; almost 1,200 Emprise units have been shipped in the ten months since its launch.
A conclusion to be made is that established vendors will tend to reject new technologies that threaten revenues from existing business models without very good reasons. Sealed canisters of drives could have put a severe dent in disk drive-based break-fix service revenues and there was no compelling reason for them to adopt the technology at all.
Xiotech's 1,200 units and Atrato's 15 are a drop in the ocean of storage array sales and won't change a thing. Perhaps when Xiotech reaches 10,000 units the other vendors will sit up and take notice. ®