chris - Fotolia
- Rich Castagna, Vice President of Editorial
What better way to start a new year than to plan and build something brand new? My pet project would be an enterprise storage system. (OK, I really had my eye on that do-it-yourself Ferrari Portofino kit, but this is a storage column after all.)
Unfortunately, my extremely limited engineering abilities and all-thumbs hands mean I won't be able to cobble my dream enterprise storage system together myself. My imagination doesn't have those limitations, however, so I'll describe it here. I want my new storage system to do everything and to do it well.
Some of the newer storage techs that have gotten a foot in the data center door over the past few years share a common theme: Storage has been too hard -- hard to buy, hard to set up and configure, hard to use and hard to maintain. Easy, the new industry byword, might be a difficult concept for some storage pros to wrap their heads around. However, today, LUNs are shunned and provisioning is no longer a three-day turnaround, but rather a menu pick for end users.
HCI to the max
Nowadays, if you can't call your storage hyper-converged or software-defined, it's probably not really storage -- at least not in the 21st-century sense. Given that, my custom-built enterprise storage system would be built around a hyper-converged infrastructure architecture, but with a couple of variations on the HCI theme.
My system would integrate the key components that make HCI, well, HCI -- storage, servers and networking. But it would have such closely integrated software-defined storage management with software-defined networking and software-defined servers, everything could be throttled up and down and configured and reconfigured on the fly, so storage performance and capacity could be rejiggered as needed. Every component -- CPU, memory, network interface card, network, whatever -- can be manipulated. That kind of endless flexibility would allow my system to morph into whatever was necessary at any given moment -- to make sophisticated decisions like what data to tier, when to tier it and what to tier it to. It will be one big whole software-defined enchilada.
My HCI system will also let servers outside the architecture access its storage resources. That way legacy gear can be part of the new world order as well. And the enterprise storage system will allow you to carve out storage regions that provide custom media configurations based on need. This means the system would accommodate all kinds of media: really fast flash; fairly fast flash; and pokey, cheap and commodious hard drives.
Cloud access would be built in -- of course! -- to allow access to multiple cloud providers for live, backup or archive data. This would be enabled by a file system that works like an automatic transmission, shifting transparently among protocols -- block, NFS, SMB or object -- as required by the apps accessing data. Seriously, users shouldn't have to go under the hood of the storage machine to get that kind of protocol flexibility. It should just happen.
I'd also add continuous data protection-based backup, which has been languishing too long on the storage sidelines. It's time for ongoing data protection that backs up data automatically to other storage systems or the cloud or Venus. Wherever it makes the most sense and recovery is easiest, just as long as you don't have to do anything but point the system in the right direction. While it really doesn't matter how it gets done, whether it's via native software or third-party apps, we definitely want to do away with proprietary formats. So we can recover data with other tools or simple copy commands.
Encryption -- in flight and at rest -- will be the default. And if I can't incorporate full-fledged security tools, the system will at least alert users when anything doesn't quite look kosher, such as off-hours access, sudden activity or unexpected access to secondary data.
Storage, manage thyself
This dream enterprise storage system would also support mega-metadata. It's an amped-up level of metadata that enables data to be smarter than us, or at least more on the ball than we typically are. So the data knows what to do with itself, what level of protection it needs based on its sensitivity or usefulness, how and when it should be tiered, who can read it and copy it, and when it should self-destruct by pressing its own delete button.
And, of course, all of this should have endless scalability to grow to those elusive n numbers of nodes or whatever -- storage, servers, network, you name it. Your starter kit might be the size of a Cracker Jack box, but it should be able to expand to Google-ish dimensions using easy, in-place upgrades to ensure you're always running on the latest technology.
Truth or fiction?
A lot of the stuff described here is actually available, so it's not all pipe dream stuff. The problem is you'd be hard pressed to find a single product that has it all. The first vendor that gets there will corner the storage market for sure. OK, maybe not the whole market; maybe just me.