In looking ahead to the promised land of cloud computing, there are not just strategic decisions to be made. Moving...
to hosted platforms requires more granular and integrated planning. Not all of these criteria for a cloud computing plan have to be met for all situations, but they are guidelines from those who have been there.
Processes, applications and/or data should be independent
Word is that cloud computing, the latest extension of virtualization, is strong and getting stronger. But experience shows that we should focus on a subset of our systems since the cloud is still in its adolescent phase and needs to mature before it supports a complex enterprise. Look to independent, standalone designs as good choices to offload to external providers.
Examples could be SharePoint applications, archived storage, virtualized desktop apps, or call center and VOIP products, most of which stand alone in how they are built and managed. Telephony is already in full swing as a Web app, from IP-based conference calls to desktop sharing via apps like GoToMeeting or Adobe's underappreciated Connect Pro Web conferencing, which include voice features. As a tie-in, existing VOIP systems should be able to extend out to the cloud if the systems are designed to scale.
On the flip side, processes that are largely dependent on each other won't provide enough return on investment due to the high cost of providing that integration across domains to the cloud. Apps that are imbedded in Active Directory -- messaging and enterprise monitoring tools, for example -- don't move easily to the cloud.
Points of integration are clearly defined
Whether it's physical integration -- creating a network portal to external devices -- or maintaining required levels of authentication, the divide between internal and external networks is unique for each service. As the specific process or application is flagged for hosted computing, it's the small- and medium-sized businesses and designers who hold the keys to success.
Bring BSAs and users to the table to lay out the methodology behind every business-critical system and find where internal security meets external open standards, as well as the bandwidth/latency considerations. Develop ways to transition from one to the other, and have these discussions before the first pilot, or your cloud may crash and burn.
A lower level of security can be accepted
Do you really need mapping software to have the same security protocols as the financial system? If not, that might open it to the cloud. The potential list grows as we dig deeper. Email has a separate security requirement from IM, although today they are often grouped together.
The holy grail of a pure-play cloud design is a network inside that only manages company user and group rights while all the systems and processes transition out to the providers. Monolithic security standards tend to pull apps inside because the individual components are visible. This is a chance to break off the pieces that are less restrictive and hold them to a different security standard.
The internal enterprise architecture is robust and well-managed
Listen to your network admins. Sniff out the DBAs for honest appraisals. Savor intelligent reporting on business application performance. Feel out the will of the staff that will support a new design. And look carefully at the reports and monitoring you depend on to measure the infrastructure.
Taking time to sense the truth about the computing environment will tell you whether or not all the parts are optimal, so that when it's time to measure performance in the cloud, you can trust your baseline information.
The Web is the best platform choice
Still, today, fat clients hog bandwidth and slow the user experience, in tandem with legacy systems that have been shoehorned into a Web front end. These often hide brittle back office problems that can't migrate off their old hardware and just won't translate into a Web-hosted environment. At this point, there is no place for applications that aren't browser-based. There are too many unknowns around latency and service levels, so keep the old client-server stuff inside. But if it's browser-based, the cloud is a great place to go.
Cost is a primary factor
Recently, the New York Times needed to place its digital archives where subscribers could search for and purchase articles. Building the storage and systems to allow that would have taken at least 24 months. Using a hosted solution, the paper converted all of its data in one week. Do we really need to break down the cost savings, or is it clear that this is the future?
Applications are current
Old apps built on the client-server design are still lunging around your network and have no place in the wild. New apps are built from the ground up (or should be) to keep processing centralized and deliver the smallest amount of data necessary across the wire...essentially designed for the Web.
For more on this approach, see the review on IT Knowledge Exchange of the book "Cloud Computing and SOA Convergence in Your Enterprise: A Step-by-Step Guide."
As recently announced, the biggest enterprise in the world, the U.S. government, is consolidating to vendor-owned data centers. That's the cloud on steroids, so the time is now to join that effort and take advantage of the latest edge that IT can give to business.
ABOUT MARK HOLT: After a 15-year climb from desktop geek to manager of complex enterprise IT projects, I joined the million-man-and-woman march of laid-off citizens. Competing with the zillions of sharp knowledge workers who are now on the street, I have been forced to learn the rules of 21st-century job hunting. I'm a resident of Virginia, a native of California by way of Kansas.
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at firstname.lastname@example.org.