How to ensure the success of a private cloud?

Written by Benoît de Pompignan, on 06 February 2018

Providing a ready-made VM via a self-service portal is a technically feasible task. However, few private cloud projects are genuine and outright successes. We will try to understand the reasons behind these partial successes.

Obtaining just a VM in less than 10 minutes rarely brings value to users. After all, what is the operational gain if it still takes 5 weeks to make it functional with DNS resolution, network openings, and user access?

The previous example clearly shows that the real added value is being able to use the machine. This implies, in addition to providing VMs, automating (and potentially virtualizing) other parts of the Information System. From experience, this can gradually touch a large part of the infrastructure services as user requirements evolve.

Value-driven management is all the more important as a project solely driven by technical aspects can ultimately prove counterproductive. I participated in a private cloud project at a major French insurance company, whose goal was to pool computing infrastructures. The principle was to automatically provision and decommission machines in real-time based on the needs of actuaries. The platform worked so well and seamlessly that actuaries used it without worrying about resource allocation. This resulted in an increase in the overall bill, even though the project was initiated to lower it by pooling infrastructure.

If this project had been driven by value, the question of expense control would naturally have come to the table. It is obvious that decision-making processes are different when resources can be provisioned in real-time than when it takes 6 months to obtain them.

Setting up a cloud also implies new practices. Typically, for simplicity, we prefer to:

  • Launch a new VM and apply an existing configuration to it rather than restart and diagnose a crashed VM.
  • Launch a VM with the latest available image rather than patch an existing VM.
  • Adjust the number of VMs in real-time rather than size the infrastructure to absorb activity spikes.

Through these new practices, the culture of automation is propagated to platform users.

It is also important to have a reliable platform to meet these new practices. In one of my experiences for a major French bank, I observed that setting up a private cloud genuinely facilitated VM provisioning. However, the service only provided VMs 70% of the time due to platform stability issues. As a result, developers rarely wanted to take the risk of deleting their VMs and sized them according to the peak loads of their applications. In the end, the cloud platform users were not unhappy: they obtained VMs more quickly than before. But with the platform's quality not up to par, it was a triple blow for the IT department:

  • It did not achieve the expected ROI due to over-sizing of the infrastructure.
  • It accumulated significant technical debt, as too many resources were used for platform support and expansion.
  • It degraded its image due to poor service quality.

Whether in terms of value or usage, the impacts of implementing a simple VM self-service portal are significant. To fully reap the benefits, it is crucial to integrate the technical approach to creating a private cloud into a more comprehensive DevOps transformation of the IT department. This transformation will require not only technical changes but also cultural, organizational, and methodological changes.