Hard Truths About Cloud Differences

8 Cloud Tools For Road Warriors

We're long into the hype cycle of cloud computing. That means clear criteria to assess and evaluate the different options are critical. Which of the many cloud approaches should medium to large enterprises take to optimize their data center operations?

Typically, the cloud is envisioned as an accessible and low-cost compute utility in the sky that's always available. Despite this lofty promise, companies will need to select and build their cloud environment carefully to avoid fracturing their computing capabilities, locking themselves into a single, higher-cost environment, diminishing their ability to differentiate themselves and gain competitive advantage -- or all three.

[ Is software as a service becoming commonplace? Salesforce: Every Developer A SaaS Vendor ]

The chart below provides a primer on the different types of cloud computing. Note the positioning of the two dominant cloud types:

-- Specialized software-as-a-service (SaaS), where the entire stack, from server to application (even version), is provided with minimal variation.

-- Generic infrastructure- and platform-as-a-service (IaaS and PaaS), where a set of server and operating versions is available with types of storage. Any compatible database, middleware or application can be installed to then run.

The two dominant cloud types

A private cloud essentially is IaaS that an enterprise builds for itself. The private cloud is the evolution of the corporate virtualized server and storage farm to a more mature instance with clearly defined service configurations, offerings and billing, as well as highly automated provisioning and management.

Another technology that affects the data center is the engineered stack. This is a further evolution of the computer appliancesthat have been available for decades -- tightly specified, designed and engineered components integrated to provide superior performance and cost.

These devices typically have been in the network, security, data warehouse and specialized compute areas. Firewalls and other security devices have long leveraged this approach, whereby generic technology -- CPU, storage, OS -- is closely integrated with special-purpose software and sold and serviced as a packaged solution. The engineered stack approach has moved into data analytics, application servers and middleware.

Over the next five years, vendors will continue to expand their SaaS and engineered stack offerings because they offer higher profit margins and more- certain long-term revenue streams. Customers will embrace SaaS for its ease of implementation and potentially variable cost, while they embrace engineered stacks for their higher performance and potentially lower costs. You should choose SaaS and engineered stacks where the business cases make sense, with the following cautions:

SaaS Caveats:

-- Be very cautious if it's a core business functionality. You could be locking away your differentiation and ultimate competitiveness.

-- Before you sign the contract, know how you will get your data back should you stop using the SaaS application.

-- Make sure you have ensured the integrity and security of your data in the application vendor's hands.

Engineered Stacks Caveats:

-- Understand where the product is in its lifecycle (older products might not provide lasting benefits).

-- Anticipate the eventual migration path as the product fades at the end of its cycle.

For both kinds of products, avoid integrating key business logic into the vendor's product. Otherwise, you'll face high migration costs once the product hits end of life or there's a more compelling alternative. There are multiple ways to ensure that your key functionality and business rules remain independent and modular outside of the vendor service package.

With these caveats in mind, you'll be successful with your decisions at the project level. But you should drive optimization at the portfolio level as well. If you're a medium-to-large enterprise, you should be driving your internal infrastructure toward a private cloud. Virtualization is just the first step. You should move to eliminate or minimize your custom configurations -- preferably to less than 20% of your server population. Next, invest in the tools, processes and engineering to automate the provisioning and management of the data center. Doing this also will improve quality of service.

Make sure that you don't shift so much of your processing to SaaS that you balkanize your own utility. Should you overreach, expect to incur heavy integration costs on subsequent initiatives, as your functionality is spread across multiple SaaS vendors' data centers. Also expect to experience performance problems as workloads operate at WAN rather than LAN speeds, and expect to lose some negotiating leverage with SaaS vendors as you lose your "insource" strength.

Nonetheless, over the next five years SaaS and engineered stacks will be a part of nearly every company's portfolio. Strong IT shops will leverage these capabilities judiciously to avoid vendor lock-in and other pitfalls, by developing their own private cloud capabilities and retaining critical intellectual property. We'll see far greater data center efficiency as we reduce custom configurations. So although the prospects are indeed cloudy, the future is bright for the thoughtful IT shop.

What changes or guidelines is your organization applying as it evaluates and deploys cloud computing? Please add your perspective in the comments section below.