Why The Cloud You Want Is Not The Cloud You Deserve

Companies need to walk before they run as they build out their own clouds.

Every enterprise wants to be Amazon these days. Tempted
by the promise of cheap, elastic cloud computing, companies are rolling
out OpenStack and other private cloud solutions, certain that they're
building a winged unicorn.

Red Hat's Alessandro Perilli

In reality, many of them are building a donkey.

That's
the message from former Gartner analyst and Red Hat general manager
Alessandro Perilli, who delivered it at the most recent Red Hat Summit.
But rather than mope in the valley of diminished expectations, Perilli
suggested a few ways businesses can grow into their most optimistic
dreams of cloud computing.

Lowering Your Sights For The Company Cloud

Enterprises
have huge visions for their private clouds, based in large part on
their experience with public "infrastructure-as-a-service" clouds like
Amazon Web Services. So when they take OpenStack or some other private
cloud technology for a spin, they assume it's going to be:

• Simple

• Cheap-ish

• Standardize and fully automated

•
"Elastic" at the application level, meaning it will automatically
expand or trim back computing resources for any given job as demand
waxes and wanes

• Resource consumption-based

• Infinitely
scalable at the infrastructure level, meaning that companies can build
out their cloud deployment as much as they want

• Minimally governed like AWS, with the ability to get work done without heavy interference from IT (and IT policies)

What they actually find, though, is a setup that is:

• Undeniably complex

• Expensive

• Only partially standardized and barely automated

• Not as elastic as they'd hoped

• Capacity constrained

•
Heavily governed by IT attempting to layer onto cloud resources a full
suite of access and financial controls, key management and encryption
and other tools

This is what the reality of enterprise cloud applications look like today, suggested Perilli:

Not surprisingly, once enterprises bump into this uncomfortable reality, the public cloud starts to look even more appealing. Turns out that it's extraordinarily difficult to build an internal data center that functions like a fully elastic cloud.

Start Small, Dream Big

Rather
than launch a fully-automated production cloud, Perilli suggests that
organizations start small, even if they're ultimately thinking big.
After all, there’s a progression of complexity as enterprises attempt to
move data center applications to a cloud environment. The more
components an app has, the more provisioning is required. The more
pieces an enterprise tries to automate, the harder its job will be.

As he argued:

Yes,
you can build a cloud in six months, but that is just the beginning.
You need to add pieces and evolve the solution, and get more mature and
add more control, and more capacity management. You don't build all
those components overnight. 

That difficulty level
goes up considerably once legacy systems are factored in, such as those
old mainframe applications that are essential to the enterprise but not
necessarily easy to integrate into cloud-based systems.

As such,
it can take a company a year to convert 10 applications to the cloud.
Given that enterprises will often have thousands of applications to
convert, the task can seem daunting.

Perilli offers a few suggestions as to how to get started without getting overwhelmed:

1. Reset your expectations

2. Select
a cloud solution that adjusts to your organization's technological
proficiency. Don't buy a vendor's bundle with six to 12 management
modules when you can hardly use one or two of those

3. Build a
solid development and test cloud first—and don't believe anyone who
tries to sell you on a "fully automated" production cloud

4.
Introduce support for scale-out application in a meaningful way, in
particular by considering a multi-tiered cloud architecture

Build With Trust

Finally, precisely
because private clouds are still technologically immature, Perilli
suggested that the most critical component of any cloud has nothing to
do with technology:

[We] need to realise that, exactly
because no solution is perfect despite all the marketing claims, what
is more important is the trust, the relationship, the credibility that
the vendor has.

In other words, building a viable enterprise cloud is a journey. So it's important to pick your travel companions wisely.