To Make the Most of Cloud Computing, Companies Are Going to Need Sherpas

Cloud computing promises simplicity to corporations: drop the data centers and plug a business directly into online resources. But outward simplicity belies the complexity of sorting through hype and the technical complexities to make cloud computing work. To do so, corporations will need a new type of specialist -- a cloud sherpa, if you will -- who can cut through the tangle and connect together online building blocks to make a virtual system that can offer real performance at the right price.

The big problem in cloud computing is that a company must assemble mysterious black boxes that offer functions, but little to no insight into how they work. IT veterans get thrown because the metrics they use to monitor performance are unavailable, and not all clouds are good for all needs, no matter what vendors want customers to think. Like mountaineers in the Himalayas guided by Tibetan sherpas, executives must rely on people who understand the territory:

• The technical landscape of who offers what
• Different routes through the landscape
• The implementation, performance, and cost implications of each choice

But straight performance results only tell you so much. In this test, Google AppEngine was the most cost-effective for small workloads, while Azure was least expensive for medium- to large-workloads. Additionally, the study leaves some big questions. How do you combine issues of storage and database performance? You might even wonder what part of Microsoft Azure the group tested, because it was unclear. To use the results, you need someone who can dig into them and translate the information to know how it could affect your operations.

Scalability is important, but how quickly and reliably will information get from the cloud to the corporate client? As the company's CIO told me last fall, Inteva, an auto manufacturing spin-off from Delphi, moved most of its IT infrastructure to cloud services but kept the manufacturing systems at its factories. The company literally had only two hours from the arrival of a car-company order to delivery of the manufactured parts.

With a cloud resource, you have no control over its operation nor the decisions that can affect performance. A vendor could make a passing decision in operating its system that could slow Inteva down unacceptably. A single Internet outage could be disastrous. So each factory had to have its own manufacturing IT system on premises.

One consulting group, CloudHarmony, has posted the results of download speed tests for a series of cloud service providers, including Amazon and IBM. This is useful data, and yet the information is limited. Downloading a file is far different from monitoring throughput on a continuing basis. A business also needs to understand how much moving data costs. At what point does speed of data delivery push make the service unacceptably expensive for the business?

The complications only increase. How do you allocate and pay for services so that you can scale operations up and down as necessary? How do you meaningfully monitor performance if you can't see directly into the black box? What business functions must stay in-house, and does the answer change as new cloud services become available?

Most IT departments can't afford the time to obsessively follow cloud vendors already on the market, analyze the trade-offs between offerings, and track every new development. Companies will need specialized guides who can help them sort through services, build systems by connecting possibly disparate vendors, and ensure that the result delivers the desired performance ... and the cost savings.

Image: Flickr user stevehicks, CC 2.0.