Interesting "lessons learned" from our internal implementation of a private cloud. Basically, it's start small and draw a line in the sand at complexity. Here's the rest of the article:
Lessons LearnedCisco IT shares the following lessons learned with other organizations planning to offer IaaS:
• Prepare early by virtualizing the server environment and implementing a wire-once server environment that eliminates the need to individually cable new servers.
• Use commercial, off-the-shelf components instead of developing the software internally. "Using Cisco Intelligent Automation for Cloud accelerated deployment from 24 months to 2 months," says Cinque. "In addition, we didn't have to train internal IT resources to maintain custom software, and we have the option to out-task management of the IaaS program to Cisco Services or a third party."
• Develop the operational model early. Map out every step for a service before you take it live.
• Know your clients and their expectations. "Early communication with customers is essential," says Jim Heil, with the CITEIS Client Engagement Team. "They will always want additional capabilities, so you have to draw a line for the first release and constantly follow up, keeping a close eye on the options from external IaaS providers."
• Start small. Make sure you can provision the simplest resource end-to-end before offering more complex resources.
• Make a simple user experience a high priority. "It doesn't matter how great the automation is if the service portal isn't easy to use," says Cinque. Hide the complexities of the offering by creating short, simple online forms.
• Calculate the TCO for the environment if you are going to implement a chargeback model.
Also I really like what they have as next steps: focus on the application. Some of this may be very obvious to my Silicon Valley start up friends ("re-write the app"), but it's positively daunting for a large enteprise.
Next StepsThe location of resources will become less important as Cisco IT's vision is to host the complete software lifecycle, including development, test, and production, in the cloud (Figure 5). This will require that applications have the intelligence to be in the right locations at the right time. Supporting a distributed architecture will also require rewriting certain applications to tolerate latency and disconnects.
Figure 5. Cisco IT Continues to Automate IaaS ProvisioningShaw-Jen Chang, Vice President of Network and Data Center Services, is committed to continually enhancing Cisco IT infrastructure services. "We will continue to improve efficiency on the PaaS layer," Chang says, "and increase cost-effectiveness when the PaaS layer is automated."