Brags about "reinventing" or "disrupting" this or that kind of technology are so common these days they might as well be white noise. But Seagate seems to have just developed a new kind of storage system for the cloud that might well be quite disruptive -- if it ever gets implemented by cloud builders.
Seagate calls it the Kinetic Open Storage platform, and it's designed as a way to enable cloud data centers to improve the way they handle storage by having the storage system itself offload as much of the processing related to storage as possible.
Here's how it works in plain English: KOS involves hard drives that use Ethernet as their sole physical interconnect: not Ethernet as in an SATA-attached drive with an Ethernet controller, but Ethernet, period. Instead of using an OS-level filesystem, data is fetched from and stored to the drives using a key/value system serialized with Google's Protocol Buffers mechanism. (The API used for data access is to be open source.) Most everything else you'd associate with cloud data center storage is pared away.
A crucial detail about this setup, Seagate claims, is how it moves many of the issues normally associated with the OS or storage-management layer -- quality of service, migrating data between drives, at-rest encryption, and so on -- to the drives themselves. By getting rid of much of the hardware associated with the traditional storage tier, you make racks denser, use less energy per unit of storage, leverage the existing data-transport fabric in the data center (Ethernet), and have storage "truly ... disaggregated from compute."
Another professed advantage to ditching all of that cruft is an increase in write performance -- up to 400 percent, according to David Chernicoff at ZDNet. Ditto any common file manipulation, such as copying or moving from one drive to another: All of that can be offloaded to the drives themselves.
A break this radical from the way storage traditionally works wouldn't come without a cost, though. Here, the price would be software development, as every piece of software that touches a file system in some way would have to be reworked to use KOS. Even with the KOS tools offered for free -- the drive simulator and developer's tools, and the KOS API itself -- the cost of such reworking would be far from trivial.
One of the reasons why drop-in replacements for existing storage systems are so appealing is because, well, they're drop-in replacements. Many of the problems still faced by cloud storage systems -- particularly write bottlenecks -- are solved either by throwing more hardware at the problem, or at the software level, by more intelligently managing data throughput. Microsoft's new release of Windows Server has intriguing new storage features in this vein, for instance.
Again, tossing out the entire storage layer as we know it is a colossal project. There isn't even a date set for when the hardware itself will be available to testers, let alone vendors or end-users. And the costs of moving to such hardware in the long run can't be ignored -- including the sunk costs of ditching so much existing legacy storage. (There's no word on if existing drives could be retrofit to use this system; I don't think it's likely.)
All this leaves a wide margin of time to see whether or not Seagate's new game-changer really will change any games. But the bare outlines of the idea alone are tantalizing.
This story, "Seagate cooks up game-changing cloud storage hardware," was originally published at InfoWorld.com. Get the first word on what the important tech news really means with the InfoWorld Tech Watch blog. For the latest developments in business technology news, follow InfoWorld.com on Twitter.
Category: Lake Natron Voyager 1
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.