You're midway through a sequence of posts describing the concept of building an "information utility" for the non-mission-critical information that's drowning every enterprise.
I did my best to lay out an initial conceptual framework here. Much of what I say here will make far more sense if you have the chance to read this first.
In this post, I want to drill down on the enabling technologies and automated operational models that have the potential of dramatically transforming the efficiency of how we store, protect, manage and leverage our enterprise information.In the next two posts, I want to explore related concepts: "control", and "choice".
Exploring Notions Of Efficiency
When you use the word "efficient" with most people, the first concept they immediately gravitate towards is acquisition costs. That's fair -- storage can be expensive stuff.
A little harder to get to -- but still there -- is the notion of operational efficiency: the labor involved in transforming technology products into a service for the business.
Then there's notions of optimization: right service level, right time, no effort required. Put differently, cheap and good shouldn't be mutually exclusive.Finally, there's the notion of adaptability: working well with what I've already got, but providing a foundation for doing things far better in the future.
For me, all of these are key aspects of "efficiency", and we'd like to see them all in our idealized "information utility".
Saving Money On Physical Storage
We all know the sad state of affairs when it comes most storage found in enterprises. Too much information stored on traditional FC drives. Poor utilization of the media for a variety of reasons. Redundant information everywhere.
Power and cooling issues. Technology and refresh issues. If you're a storage administrator, you live in this world, and you know it's not ideal.The good news?
Storage vendors have been dumping enormous amounts of R+D into this space, and we have an entirely new set of tools that directly attack the "cost of physical storage" issue in an eye opening way.
Intelligently combining enterprise flash and big SATA drives can deliver 20-40% physical cost savings. Not only that, you're in the sweet spot of the technology roadmap where all the big cost reductions are being forecast.
Using virtual and thin provisioning techniques with various forms of reclaim can take another 10-40%, depending on your environment.
Apply data reduction techniques, such as advanced dedupe, compression and single-instancing -- there goes another 10-50% of physical media -- maybe more if you're lucky.
Start to spin down the drives you're not using, and *bang* there goes a big hunk of power and cooling costs as well.
Put all of these goodies together -- and make them usable -- and you've got the potential for a 60-90% cost reduction for acquisition, power and cooling associated with physical storage media -- at least, compared to the typical practices associated with using FC drives.
Even if we're being conservative, there's at least a 50% reduction there for most people just by using a combination of the newer storage technologies coming into the marketplace. Maybe more. And this stuff *works*.And what IT professional wouldn't like to cut their costs in half?
But we can't do that at the expense of making things more difficult.
Saving Effort On Physical Storage
I spend a lot of time working with larger IT organizations, and I learn interesting things along the way.
One thing I've started to do is find all the people who touch non-mission-critical storage in different ways.
There are people who do forecasting and realize when it's time to buy more.
There are people who select the technology, negotiate with vendors, and buy the stuff.
There are people who take the new stuff, get it on the floor, and make it usable to others.
There are people who monitor service levels and capacities, and report out.
There are people who work with finance and business units trying to get this stuff paid for.
There are people who respond to problems around performance, or availability.
There are people who backup, recover and archive the information.
There are people who think about disaster recovery and business continuity.
There are people who worry about security, compliance and auditability.
If you're the proud owner of hundreds of terabytes, or perhaps petabytes -- how many people in *your* organization touch this stuff? In a decent-sized IT organization, I can usually find 30-50 people (sometimes much more) that spend all or part of their time on these tasks.
Now, let's take another view at people who *use* this stuff as a service.
There are application and server people who need this stuff to do their job.
There are desktop users who need a place to store their stuff.
There are business analytics users who like big playgrounds.
There are people developing new applications for the business.
There are legal groups doing e-discovery.
There are finance groups looking for records management.
The list goes on, and on, and on. What does your organizational user list look like?
The reason I made the second list is simple: it's far too easy to think simply in terms of what the storage team wants. Our ideal "information utility" has to meet the needs of everyone who uses it, doesn't it?
When we speak of "saving effort", we want to make sure we have the complete picture, and not just a subset of it!
So, what new technologies are on the horizon that make big chunks of the IT-related effort go away?
Glad you asked.
If we're running this idealized "information utility", I should be presented with choices when it looks like I'm running out of something.
Here are your choices to change policies to increase capacity (or performance, or protection), and here are the likely impacts. Or, if you need new stuff, here's the amount and type you need, and here's where it should go. Basically, information to make intelligent decisions.
If I do decide to buy new stuff, it should come out of the box and set up with a minimum of fuss. It should ideally be auto-discovered, auto-provisioned, and re-balanced automatically.
Dynamic policies should automatically be pushing unused information down to the cheapest pragmatic tier, unless there's an exception. Backup and archiving should be on autopilot.
If some subset of information needs special handling (e.g. performance, availability, disaster recovery, compliance, security), that should be as easy as selecting a few radio buttons, and it's just done. No drama, no fuss.
If I'm responsible for providing this information utility service, I ought to be able to be alerted if something is trending out of spec, rather than having to continually monitor things. When it's time to drill down into the plumbing, I should be able to do so quickly and in context. And if I have to report out to somebody about something, that should be pretty easy as well.
That sort of model seems to work pretty well with power, phones, etc. -- why should providing an enterprise information utility be too much different?
The vast majority of "making capacity available" ought to be integrated with something else, and not a standalone administrative function. Provisioning a virtual machine? Storage is part of the activity. Writing a new generation application? Here are your APIs. Need a place to put your stuff? Here's a self-service portal.
Efficiency As In It Works With What I've Already Got
One of the painful realities is that, in every enterprise, there's a pile of stuff already on the floor. Sure, we'd all like the luxury of sweeping it all alway, and starting with a blank piece of paper, but that's going to be the exception, and rarely the rule.
If you're like most enterprises, there are big piles of non-tier-1 storage to be found everywhere: file servers, iSCSI stuff, backup stuff, archival stuff, scratch stuff, even some FC stuff. Has to work with all of that, right?
I've got information being generated on tier 1 storage, it should seamlessly find its way into the information utility space just as fast as it can safely get there. And, if I want to use an external cloud storage provider, I should be able to easily ship stuff off to them as well.
From a pure technology perspective, this is achievable through orchestration and integration. Sure, purpose-built array products that implement a better information utility are certainly attractive, but that shouldn't be table stakes to play.
Final Thoughts
Efficient means more than cheap hardware, although that's always nice.
And, as we think of our idealized information utility, we ought to attempt to expand our notions of "efficient" to think in terms of how utility providers think about the concept.
Next up -- concepts of control.
Your thoughtful comments are welcome as always.
Comments