- Blake

-Seagate isn’t alone: nurturing innovation inside a large company is a general problem that companies get wrong more often than right. It is the nature of Seagate’s errors that suggests there is a systemic corporate culture/cognition issue specific to Seagate. It would be fun to work with them on it. In the meantime, small companies remain a fruitful source of innovation – thank goodness!

– will flash SSDs ever be profitable? Good question. The flash manufacturers are running a process factory in an industry with significant seasonal demand – which SSDs hope to correct. The skills needed to develop and market flash SSDs may be dissimilar enough so that we end up with something like the disk/array split – where disk vendors aren’t in the array business – and flash vendors aren’t in the SSD business.

-There are at least 2 areas of innovation open to drive vendors who don’t want to compete with their customers: new kinds of disks – Cold Archive Disk; and packaging disks into “super disks” – the Xiotech ISE. That little Xiotech is able to sell the latter when big Seagate couldn’t/wouldn’t is why it appears Seagate has some systemic problem.

– no audited SSD TPC numbers? Interesting. What is the problem?

– SanDisk did a press release on the SD Worm card, but they aren’t selling them on the company store. How about an update SanDisk?

Robin

http://www.sandisk.com/Corporate/PressRoom/PressReleases/PressRelease.aspx?ID=4353

In a lot of real-world workloads, IOPs don’t get any near the ridiculous levels of flat-out TPC’C’ I/O levels. However, latency does tend to be a really serious issue as many such workloads have limited multi-threading capabilities (whilst TPC’C’ scales very well with multi-threading). It’s for these reasons that arrays with write caching can make big differences on performance – it’s not total IOPs, it’s latency that limits throughput on many real-world applications.

Physical disks have latency limits that are very unlikely to be fundamentally improved, and caching can only go so far as it gets into laws of diminishing returns. What is required is a major breakthrough to deal with latency issues, especially on random reads (it’s often easier to deal with latency on writes through NV caching and write-back).

Many published benchmarks have this problem – they major on total throughput on heavily multi-threaded workloads rather than performance on thread-bound applications.

I want a drive that I can put my photo on and put in a safety deposit box for ten years and still be able to access my photos with a minimal error rate.

Of course there are more applications than digital photography, and the manufacturers who jump on this idea will create a whole new market segment.

Why on earth do I want to buy AIT-5 tapes at $45 per 400GB when a HD is cheaper and I don’t need a $6000 library to read it and make it useful for backup?

The average photographer doesn’t need 100MB/s throughput, they need security, longevity, and robustness.

Capacity isn’t even that big an issue as most home users are only shooting 10-20GB of family photos a year. You could be selling them one drive every year for their archive needs.

Like a number of forecasters, I’ve done the calculations on when flash will beat 15K and 7200RPM drives, etc, on $/GB and other metrics. 15K is soon, 7200RPM not so soon. With flash excelling so rapidly, I’ve wondered why the vendors aren’t focusing on creating a new class of enterprise drive: the enterprise hybrid. This is a 5400 or 7200RPM drive, crafted expressly for density and power conservation, coupled with a flash drive (ala the laptop, but for enterprise purposes). The flash-hybrid aspects of the drive would be to increase r/w IOPS: a giant MRU cache, if you will. I ought to only need a few GB of flash to be able to achieve significant performance benefits. This would be a drive with tier-1 workload properties, but tier-2 densities.

I would think the win here would be enormous: it would create a margin segment out of the low margin drives, something the manufacturers are going to be desperately needing when the 15K drives go buh-bye. Which they will: very soon now.

Joe Kraska
San Diego CA
USA

His last go-round at Seagate ended with the sale of the software assets to Veritas and then taking the company private. It will be interesting to see how this turns out.

It’ll be interesting to see which market tanks more this year, NAND flash or HDD. My money is on NAND flash – overall semiconductor is already tanking at >20% sales decline.

Seagate already does sell direct to consumer, e.g. FreeAgent, Maxtor One Touch.

As long as the demand for storage capacity increases, HDDs should do fine (due to its capacity lead over NAND flash). If demand starts to level off, then I’d be really concerned.

Next question: if Seagate must succeed as some kind of solid state storage technology provider, can they do it without manufacturing scale? I don’t personally see how Seagate could compete successfully (=profitably) against the Intel-Microns and Samsungs of the world in a traditional semiconductor-based solid state technology like NAND Flash. The value-add in controller architecture is real but slim and fleeting, surely not enough to compensate for lack of manufacturing ownership and scale. Anybody put numbers and statistics to an analysis like this?

No wonder Seagate has a confused SSD strategy. I don’t know anything about Seagate’s inner workings, but I have to wonder if this was a major thorn of disagreement between the executive team and the board.

I wonder if you are looking for change in the wrong place here Robin. A huge chunk of Seagate’s business is to OEMs and ODMs so they should (in theory) be listening to those folks about what they would like to do, but your exhortations (which are spot on for new opportunities) should probably land in the inboxes of the folks who deliver and service the end user products.

Look at the NetApp S series. Here was a box that might have made it into the home with NetApp RAID/WAFL goodness if NetApp could have figured out a way to let consumers plug in their own disks of their own capacities. Instead you’ve got the DrRobo or the ReadyNAS trying for that market.

Another product, something like just the time control properties of a digital video recorder (DVR) (Tivo patent sharks not withstanding) where you replace that time shifting marvel, the VHS VCR with an inexpensive DVR for all the folks who don’t have cable/satellite etc. But where is JVC or Panasonic in that space?

Where is the Apple “media box” that sits in your home and syncs automatically to the iTunes store for everyone in the house, can play new music ala Pandora and even a retiree could operate?

The next 10 years will find storage appearing in places it has never been (like processors did in the 90′s) and yes Seagate could get serious there, buy a flash company (SanDisk would be a good choice) and continue as the premier source of storage “widgets” to the ODM / OEM community. They have to be careful about trying to move up the value chain (which is an easier way to increase profit than improving operations) because doing so threatens the very innovators who are going to discover the next big demand for storage gizmos.

Bottom line, Seagate could win by re-dedicating itself to being the the top supplier of storage components to the OEM/ODM community and innovoating around ways to do that more cost effectively than their competitors. If they do that they have to get serious about Flash as well rather than trying to keep it out of the market.

–Chuck