Over time computing does get easier. For example, it took Earthlink founder Sky Dayton eighty hours to connect to the Internet the first time he tried. Nowadays I'm not sure you could buy a computer without access built in.
Take it out of the box, plug in the cables, turn it on, and it will hunt down the connection on its own. Further, now you can hot-plug in a new peripheral and the operating system automatically locates it and adjusts the necessary settings.
Other aspects, however, have not yet arrived at that level of simplicity. Such is the case with storage area networks (SAN). Although the technology has been around for years, at a recent Storage Networking Industry Association (SNIA) conference, twenty experts from the equipment vendors spent the entire night trying to get all the pieces of the state-of-the-art SAN demo to work together properly. The individual components are superb, but they just don't play well together.
"The industry is fractured and focuses on point products," says Yogesh Gupta, Chief Technology Officer for Islandia, NY-based Computer Associates, Inc. "A high level of abstraction is needed to hide the plumbing and simplify the management view."http://o1.qnsr.com/log/p.gif?;n=203;c=204655439;s=10655;x=7936;f=201806121855330;u=j;z=TIMESTAMP;a=20400368;e=i
The main problem with enterprise storage systems is that they are, essentially, one-off rather than commodity items. It's similar to the early years of the automotive industry. Automobiles are complex items containing tens of thousands of parts, and they used to be custom-built the way many storage systems are now. But today if you want a new car, you just walk into a showroom, go for a test drive, sign the paperwork, and drive it home that same afternoon.
Although the components under the hood vary greatly from one car to another, essential aspects of the human interface are similar. Although it may take a while to learn the intricacies of the stereo system, you don't need to attend a four-day workshop to learn how to drive it.
Now, imagine if automobiles were built like storage systems. You would hire a few automotive engineering consultants, specify basic parameters — number of passengers, cargo space, acceleration, mileage, etc. — and they would then try to come up with a mix of components that meet those requirements. After a few months, the sample parts would arrive. You would then hire a mechanic to assemble them and next spend another couple months testing it out and making adjustments to ensure all the parts work together. If it does what you want – and after a bit more tweaking – you can finally start driving it around.
True, there are significant differences between a SAN and a SUV, but that doesn't mean that storage needs to be as complex an issue as it is currently.
Black Box Storage
CA's Gupta attempted to define what it would take to simplify storage architectures, management, and usage. He laid out seven aspects to turn storage into a "black box" utility:
- Create Self-Describing Storage Objects – This simplifies administration by allowing devices to be managed based upon classes and subclasses. Object types would include controllers, the devices or volumes, the fabric elements, file system objects, and block objects.
- Establish Class-Centric Policies – Policies that govern the storage QoS (Quality of Service), including where something is stored, the transport characteristics, the availability level, security attributes, and accessibility. The policies also control aspects such as backup, restoration, and replication.
- Improve Storage Security – Since most storage is now networked, this exposes it to new types of vulnerabilities. Most security solutions are system-centric, so administrators need to establish ones that are network-centric. Companies should use the new Advanced Encryption Standards (AES), such as the Rijndael algorithm adopted by the National Institute of Standards and Technology (NIST).
- Standards and Instrumentation – This includes the Storage Networking Industry Association's Storage Management Initiative Standard (SMI-S), which is designed as a bridge between the storage devices and the management tools.
- Virtualization – Separating out logical storage from the physical devices it resides on.
- Automation – Some vendors already support automated provisioning, but configuration, self-healing, and recovery also need to be included.
- Single Point of Management – It does no good to have separate management interfaces for each different vendor. Standardizing object classes will make it easier to unify the monitoring and management of heterogeneous devices.
Okay, so all these aspects look fine on paper, but how does it all translate into in the data center? Essentially, what it represents is a shift of attention away from hardware limitations and peculiarities, and onto servicing business needs. Instead of having to worry about the nuts and bolts of which partition of which disk in which array holds some bit of data, and how to link that data with the right application or user, the IT staff are free to establish and maintain storage policies that meet the organization's goals.
Sure, they will still be able to drill down into all these details as needed, but generally speaking, the management software now handles all the minutiae, not the staff. It also means you are no longer locked into a particular vendor since the products operate on common standards. As easily as you flick a light switch or turn the car's ignition key, the system works without you having to know how or why.
So will we ever reach that point? There's no question that numerous details still need resolution, but we have standardized everything from CD formats to shoe sizes — there's no reason we can't achieve the same with storage.
Feature courtesy of Enterprise IT Planet.
Back to Enterprise Storage Forum