Building Choice Into Our Information Utility -- Part 4

You're deep into a series of posts where I attempt to lay the foundations of an idealized "information utility" for most enterprises.

The idea is simple: the vast majority of enterprise information isn't particularly mission critical, and we should start thinking in terms of an internal "information utility" that's modeled after phones, power, etc.

Before you dive into this one, I'd recommend that you spend a moment reading this post, this post and this post.

Ready?

Notions Of Choice And Our Information Utility

Talk about "choice" with IT infrastructure specialists, and it's a hot topic.  They strongly dislike being locked into a single vendor or a single approach, and for very good reasons.

Sure, they may decide to prefer a single-vendor approach for one reason or another, but they always want the flexibility of doing something different if the situation changes.  Makes sense. There's also the practical side of this -- there's a bunch of stuff already bought and paid for.  No way anyone's going to be able to walk away from that easily.

Choice also has a lot to do with how things might change going forward.

For example, if there are external storage service providers I want to work with (faster/better/cheaper/more protected), that's going to be an interesting choice I'd like to have open going forward.

Or, perhaps we want to move from a distributed model to a centralized one -- or vice versa.  That's a useful choice to keep open.

Not to mention, storage technology is moving at a breathtaking pace these days.  I'd want to keep my options open as new stuff comes to market -- better media, better security, better dedupe, etc.

For IT infrastructure professionals, choice is good.  But if we really think about it, there are other forms of choice -- and implied flexibility -- that we might find attractive going forward.

Choice Of Access Model

The storage industry started -- way back when -- with a block-oriented model and LUNs.  We then all got more comfortable with using filesystems (and LUNs emulated with filesystems!) as an access model for more of information.

This, in turn, is giving way to an object-oriented model accessed by APIs -- REST, SOAP, XAM, etc.  More metadata == more services that can be provided by our "information utility" service we're envisioning.  See an interesting preview here.

Needless to say, our information utility should support them all, using the same set of underlying resources and operational models.

Choice Of Protection and Security Model

There's lots of different ways to protect information -- all of which trade off costs for level of protection.

You may want CDP (continuous data protection) if you want to create an infinite set of logged versions that can be played back and forth like a tape.  You may be comfortable with the once-a-day or once-a-week model.  Imagine a "protection dial" moving back and forth, depending on your needs.

You may need an audit trail of who changed something, or who accessed something.  Or not.  More choices.  Sometimes information needs to be a little secure, sometimes it needs to be exceptionally secure.  More choices again.

Sometime's it's OK to have information in one location.  Sometimes two is the answer.  Sometimes you'd like certain information spread very wide indeed to make *absolutely sure* it's recoverable regardless of circumstances.  Another imaginary "distribution dial", moving back and forth, depending on your needs.

Again, none of these decisions should have to be made ahead of time.  They're dynamic choices you'd like to make, when you need them, and with a minimum of fuss and expense.

Choice of Responsibility Model

One of the perennial debates in IT is just how much control should be put in user's hands, and how much should be in IT's?

The natural tendency is to try and idiot-proof user responsibilities (often with just cause), but -- there's no denying that, over time, the secular trend is to put users more in control of their IT resources.

This implies that any proposed information utility ought to have a lot of choices as to how much the user is responsible for, and how much IT is responsible for, since there's never going to be a uniform and static boundary.

As different user groups prove that they know how to use a resource, they get more responsibility.

Choice Of Extending Functionality

It strikes most people at some point that, once you have a big pile of information in one (logical) place, maybe there's some useful things you'd like to do with it?

Enterprise search comes to mind, as an example.  Or perhaps adding some metadata to create domain-specific repositories around customers or projects.  Supporting e-discovery for the legal guys?  Or maybe (finally) implementing a records management policy that includes eventual deletion?

All of these are interesting scenarios to contemplate, but none of them can be easily accomplished without having the majority of information under some sort of logical control, and not stovepiped in dozens (or hundreds) of different storage puddles.

None of these may be in your immediate future, but wouldn't you like to have the choice?

Choice Of Consumption Model

If you're an enterprise IT organization, you'd probably appreciate some choices on how you pay for your information utility.
 
Would you like to buy the assets and run them yourself?  Would you like to buy the asset, but contract for the operation?  Would you prefer a "by the drink" model, using resources and people within your data center?  A hosted model?  Any other variant?

Put differently, we shouldn't assume that everyone wants to buy this stuff and run it themselves.

At EMC corporate, we use our fair share of copy machines and printers.  For a storage company, you'd be surprised on how much paper is sloshing around :-)  For years, there's been a bunch of Xerox people who take care of it all.  I think we look at this stuff entirely as opex (pay for what you use), rather than capex and internal headcount.

Should our ideal information utility be all that different, if we choose?

Choice Is Good

When we think about our idealized information utility, we all want choice and flexibility.  But I would argue that our notions of choice need to evolve beyond mere multi-vendor interoperability.

I'll share a few final thoughts in the next post.

Thoughtful comments welcome as always!

Comments

[32gb ssd]

This paper describes the central role that personal workstations were foreseen to play in future generation of information systems. The particular project, the Information Utility, was meant to form the foundation for libraries of the future. The paper describes the components and layered structure of a nationwide information utility. The focal point of this information system is the personal workstation. Intimately tied to this is the local computing environment, supported by powerful microcomputers dedicated as servers in the local environment and supporting workstations. Relationships between the workstation, its local environment and regional/national environments are also described.