According to Piston Cloud Computing's CTO, the rate at which his customer's pilot projects turn into production private clouds is pretty typical of most OpenStack-based providers and it's pretty low.
"Roughly for every 20 pilot projects we open up, we see one of them make it into production," says Josh McKenty, who prior to founding the Seattle-based Piston, worked at NASA and served as a technical lead on the project that evolved into OpenStack, one of three open source infrastructure as a service (IaaS) platforms vying for enterprise attention.
So why do the other 19 drop off? Good question, admits McKenty.
According to analysts, developers and cloud practitioners, the answer turns on miscommunication (between corporate IT department and their potential "customers"), dependencies (on beloved features of gear sitting in the data center or on the network) and unruliness (of applications not built to run on the cloud at inception).
In an effort to increase the chances of Network World readers landing a successful private cloud deployment, these experts offer up some tips on how to avoid the typical pitfalls.
1. Get out of the data center and talk to all business influencers.
This will help IT execs get a better idea of how the private cloud will eventually be used, advises Marten Mickos, CEO of Eucalyptus, an open source private cloud provider with close ties to Amazon's public cloud scheme.
Micros tells the story of one large customer doing business in the tech industry which built a private cloud to support dev/test across the company. "Because they didn't fully understand how much their customers' were going to consume in the private cloud, they were experiencing 50% growth per quarter," Micros says. The IT team had based the size of the private cloud on current testing levels, but the growth was seen because the users realized how easy it was to provision and run tests, so they ended up wanting to do more testing than before.
2. Set user expectations.
"Many folks read about the cost savings cloud automation can bring," says Jim O'Neill, CIO of HubSpot, and a self-proclaimed "lover of the public cloud commodity mindset. HubSpot is a supplier of cloud-based marketing services which plug into Amazon's public cloud and Rackspace's managed private cloud infrastructure.
"So you have to be realistic with your developers and business users about paying for good QoS. Even with all that they've read about the economics of the cloud, you have to make them fully understand that they can't expect five-nines uptime on a three-nines budget," O'Neill says.
The economics of private cloud vary based on the type of workload you want to run there, explains Forrester analyst Lauren Nelson. "Basic compute is cheap in the public cloud if the application is a variable one and it is written to run there. But if you've got an application that is very compute intensive or requires non-standard virtual machines that require increased I/O or memory, it is certainly more economical to run that in a private cloud setting," Nelson says.
Sign up for MIS Asia eNewsletters.