Using automation, self-service provisioning to build a private cloud

IT pros who understand the benefits of cloud computing are assessing if they can build a private cloud behind closed doors for additional security. Incorporating cloud automation, self-service provisioning and integrating IT management tools will help make their private cloud project a success.

With a private cloud infrastructure, an organisation can maximise the use of IT resources and save costs while introducing self-service capabilities and empowering end users.

A private cloud makes IT more efficient and frees up IT pros for more strategic tasks than just managing servers. Another popular use for a private cloud is to host testing and development applications, according to TechTarget’s UK and Europe Virtualisation Decisions survey of 2011.

Self-service provisioning and cloud automation capabilities
Server virtualisation is the key ingredient for building a private cloud. A pool of infrastructure resources provide the elastic characteristics of a cloud service, so it’s important to have a virtual estate which has dynamic resource balancing and management.

To deliver private cloud services, IT requires end-to-end hands-off server build automation capabilities, with the build automation process going beyond just a basic OS template build.

When offering a server workload through a portal -- whether it’s a basic Web server or a more complex environment such as a middleware stack or database -- the server provisioning request must be delivered to requesters as advertised. 

This level of automation requires IT admins to either compile a bespoke custom-built script using commands such as apt-get to deploy required services or use a third-party product that delivers this functionality.

Bespoke scripts are more automated to a service layer, such as the Web service layer. It can be time-consuming, however, to deliver an automated build for applications that are only installed on an ad hoc basis. 

IT teams must review whether automating ad hoc workloads has any time-reduction benefits beyond just giving basic access for app/dev teams to manually install components and apps. By using inventory tools, they can still ensure that this activity is monitored without having to automate everything.

But self-service provisioning has its pitfalls. A self-service-driven environment needs careful monitoring and IT control in several areas. First, because standardised server workloads are key characteristics of cloud, build revision control is required to ensure that builds are in control. This requires IT teams to review each requirement for an application and any prerequisites.

In addition, they need to set up a process to ensure that requests for build changes -- through either application updates or developer changes -- are correctly filtered, tested and then farmed into production.

IT developers will certainly like the level of control and speed at which they can spin up machines, but if that’s not controlled with a governance policy, it will soon turn into another sprawl scenario. 

They must ensure that when offering services, they have full lifecycle options for that virtual machine (VM). Whether it’s for 10 days or 10 years, they’ll need that level of information to ensure that the private cloud does not become over-allocated.

They must also ensure that the self-service tool sets are capable of metering the software and OS licensing allocation for any deployed VMs. If not, it may become a large problem when they have users and development teams spinning up multiple VMs and putting IT at risk of being under-licensed for apps and operating systems.

Integrating IT service management tool sets into the private cloud
A private cloud typically includes a “shop front” portfolio of services that internal customers can self-provision. This is known as the self-service portal. An underlying IT service catalog -- which is a library of all service offerings available to IT customers -- serves this portal.

Another requirement of a private cloud is to integrate service management tool sets to collect and monitor requests made from the self-service portal. This gives IT the ability to monitor requests made by customers and, more importantly, to approve or disapprove any requests made for server builds. 

Integration with existing service management tool sets is a basic requirement to ensure that IT service managers aren’t reinventing the wheel. It also makes IT capable of monitoring the trends and growth in deployments.

From a monitoring perspective, admins will need as much available metadata as possible about the number of requests for new VMs. To ensure a satisfactory cloud infrastructure, they also need information on how the shared pool of resources is being used.

When offering a private cloud service, project teams and developers no longer procure dedicated hardware resources. This means IT must ensure that resources do not become constrained and that hardware resources are readily available for future requests.

To ensure the infrastructure is capable of facilitating workloads, IT must set an agreed-upon threshold for infrastructure limits before more resources are added to grow the private cloud.

With the theoretical reduction in staffing required to perform labourious tasks such as VM deployment and decommissioning, IT should be able to monitor resources with monitoring tool sets. Not only must infrastructure monitoring cover virtualisation hosts, it also needs to keep an eye on storage, networking and other dependent infrastructure services.

If IT fails to proactively ensure that resources are available within a private cloud, it runs a risk of not being able to satisfy the business requirements, and this will cause business stakeholders to question the worth of a private cloud.

Cost metrics
Building an agreed-upon cost model for services is imperative for private cloud success. Cost metrics should target additional infrastructure such as storage, network and memory. A number of cost options exist, and IT can either agree on a price model for these services individually or collectively. Certain models will benefit selected use cases, and some won’t.

Here are two examples of how you could charge for services in a private cloud:
- a monthly charge for a VM and supporting hardware, software and operational management services;
- a “pay-as-you-go charge”, which works well for development and testing scenarios.

Of course, developing a new cost model is never easy; there is no one size fits all. If a business is hot on accounting and budgeting already, then executives will understand and fully be on board with a cost metric assigned to the use of a cloud resource. 

However, some IT pros may need to justify the change to business heads. Two benefits are more precise -- budgeting predictions for future requirements to prevent under-budgeting; and the integration of unit pricing into ERP systems to reduce paperwork and cross-charge reviews.

It is also important to ensure that IT can recoup costs for future hardware requirements within the private cloud.

Measuring success of a private cloud project
One way to justify building a private cloud is to measure the reductions of total cost of ownership. By nature, private cloud is characterised by highly used underlying hardware resources, so instead of focusing on the old server consolidation ratio, admins can highlight the savings made from indirect reductions in IT resources.

Indirect reductions come in different forms. One example is overhead reductions in day-to-day IT operations. Simply put, self-service provisioning means that the IT staff will be put to better use than just deploying servers. Building servers for projects or development teams is a resource-hungry and time-consuming activity, and empowering the business users to request servers themselves allows IT staff to focus on more proactive and forward-thinking activity.

User satisfaction and a rise in demand for services are other signs that the private cloud project is valuable and efficient. Demonstrating these improvements can help justify the time and money spent on implementing automation and self-service provisioning and on integrating management tool sets as part of a private cloud build-out.  

Daniel Eason is an infrastructure architect at a multinational company and is based in the UK.