Read Part 1.
In this three-part series, we look at what to consider in the early phases of a private cloud project. We already talked about virtualization and automation. But, cloud is much more than that and it’s important to have a close look at self-service and chargeback too.
To increase speed in IT, private clouds use self-service portals. This way, employees can request resources for the time they need.
In my experience, the self-service portal is really about three aspects:
- Users feeling at home
The first thing a user does is to authenticate with its corporate login. Thus it’s important to define how your self-service portal will integrate with the corporate directory. Of course, the portal also needs to be easy to use and should be rebranded if the need arises.
- An extensible portal
Your first private cloud portal will certainly be the first building block of a global cloud portal for the whole company. It should be possible to extend it toward:
- New types of hypervisors (from KVM to z/VM)
- Approval processes made simple enough
First, you have to define offerings and roles: what is offered to users, who can see what and who can request what.
When that is done the real work begins: you need to define approval processes. Here almost anything is possible: double validations, financial approvals, management approvals, and others. But when it comes to using a very complicated process for some simple tasks such as installing software on virtual machines, you want something simple enough and usable. Take the time to check whether your processes can be simplified for the same result.
Chargeback is key to the cloud, but depending on your priorities it can be used for various purposes:
- To enable IT to become more of a resource center than a cost center
- To encourage optimization of the IT by transferring the financial responsibility of provisioning to the self-service user
- To make easier and much more precise audits and financial analysis of the IT
The challenge when implementing chargeback is to get the usage data. After that the processing (rating and accounting) is not specific to cloud.
The first step to implement chargeback is, for each metric (CPU, RAM, disk, and so on), to decide with all the actors, from finance to IT, which data will be used and from which system:
- Is it the reservations from the self-service portal (if a user asks for 2 vCPUs will the user be billed for 2 vCPUs regardless of what his workload actually consumes)?
- Is it the usage acquired through an external metering tool?
The second step is to make sure that when you extract usage data you get enough information to know which department to charge.
The third step is to agree on the frequency the data is generated: are we counting CPUHours, CPUDays?
Eventually, after the rating and accounting are done, you should plan for an interface with your own corporate accounting system so that you can bill whoever has to be billed for the cloud services.
The last part of this series is about monitoring, compliance, backup, and security.