WavebreakMediaMicro - Fotolia
Software-defined was hands down the most overworked, overhyped, overused -- and abused -- descriptor in the vocabulary of data storage systems in 2017. And with a blizzard of terminology, including cloud, internet of things, big data, AI and machine learning showering down on us, that says a lot.
Software-defined stood out because it's perhaps one of the least accurate product descriptions among all vendor chatter. Consider the source of all this prattle about software-defined systems.
Yup, you guessed it. The most vocal vendors seem to be purveyors of traditional storage and server hardware. While seemingly working at cross-purposes to their own interests, they're nonetheless recasting themselves as software-defined systems vendors.
There are a lot of problems with that proposition, but two biggies really stick out.
Breaking up is hard to do
First, it's not like the old hardware vendors have quit selling hardware. For the most part, you can't go to the checkout counter at NetApp or Hitachi or most of the other big-time hardware vendors and say, "Hold the hardware. I'll just take the software, please." Chances are, you can only buy their software-defined systems on their hardware. Hmm, what's wrong with this picture, and how is it all that different from the pre-software-defined days?
The biggest sticking point with the hardware-vendor-turned-software-defined-vendor fallacy is that everything in computing has been software-defined since the abacus. It has never been just software or just hardware, and to try to separate those two seems, well, kind of lame.
The proof is most vendors aren't actually separating them. You can't. Even if you're running your own software in the cloud, you might not see it, but there's hardware there. You might not own it, but you're certainly renting it.
Of course, the whole software-defined thing is a key element of the hyper-converged infrastructure (HCI) craze. And, really, it isn't so hard to get crazy about HCI. HCI kits are pretty neat packages, closer to plug-and-play data centers than we've ever been. Is it time to grow your server and storage environment? Just pop in another node, no muss, no fuss. HCI is the ultimate in scalability -- well, usually.
Scaling up by adding new chunks of server plus storage certainly seems like a convenient building-block approach to accommodate growth. But what if you only need another couple of cores to crank through a key app, or maybe your database is simply thirsty for more storage? Some software-defined systems force you to buy new nodes that add at least some of both, even if you only need the compute power or the storage capacity. In those cases, it isn't just the software that's doing the defining. The hardware is playing a pretty big part, too.
Of course, the software-defined gang's real message hasn't been you can do away with hardware. Ostensibly, the point is standard hardware components are OK, and proprietary bits of hardware are a no-no. Theoretically, you can buy software and run it on the hardware of your choice, as long as that hardware meets the software's requirements.
But few companies actually do that, and fewer vendors want you to do that. They want to make sure their software runs as well as it possibly can. To ensure that, many of those vendors do the software-hardware integration for you, stretching the software-defined moniker a bit.
A handful of HCI vendors are all-in for supplying HCI software, but not so enthusiastic about seeing their software chug along on commercial off-the-shelf (COTS) hardware. That's why you see some HCI offerings include "specialized" nodes that may provide only additional compute power or just storage.
Proprietary may not be a problem
One reason for shared, networked storage is that users and vendors realized building storage-only systems presented an opportunity to create storage-specific hardware that would outperform what was commonly available from the server world. That's still true.
Slowly, this idea that hardware isn't that important is starting to melt away, however. For example, iXsystems, a server and storage company that's been around for years and is perhaps best-known for its FreeNAS products, touts its software, but doesn't pull any punches when it declares it's a hardware company.
I'm betting we'll see something of a correction in the software-defined systems market in 2018, with more vendors opting to offer more than just off-the-shelf hardware in their HCI and shared storage products. The software-on-COTS options will still be featured on IT infrastructure menus, but more hardware-centric choices will become appetizing alternatives.
- Software Defining Secondary Storage –Quest
- IDC: The Economics of Software-Defined Storage –Red Hat
- Software-Defined Storage for Backup and Recovery –Hedvig Inc
- Making Software-defined Storage Easy –Red Hat