The second phase of the cloud project took place from 2009 to 2011. This phase involved updating applications, specifically taking business apps, such as those used in sales and marketing, and hosting them on a single platform via VMware's vSphere cloud computing OS and EMC's own vBlock converged infrastructure platform. Today, the company is hosting 75% of its business apps in the private cloud, he said.
Additionally, EMC's 8,000 servers are 90% virtualized, and 95% of its mission critical apps run partially or fully on virtual machines, he said.
The project saved EMC $112 million on capital expenditures and an additional $66 million on operational expenses, Mirchandani said.
"Performance is blazing," Mirchandini said, adding that the private cloud has five times more scalability over the previous architecture. As a result, the company retired 65 legacy applications, moving from 450 point-to-point interfaces down to 125.
The third and last phase of the cloud project, currently underway, is to deploy IT as a service, meaning new services are automated and placed into a catalog from which business units can choose.
Critical to rolling out IT as a service is creating a baseline and then benchmarking advancements as the project moves forward. This step helps the IT group know what progress it has made and to understand the savings that result from those advancements.
While EMC hasn't completed its IT-as-a-service project, Mirchandani can already see how it will reduce costs and management overhead. If a business unit wants a new app, an additional server or more storage capacity, it simply chooses it from the service catalog.
"IT almost plays a passive role rather than an active role in this," Mirchandani said. "The moment you expose the cost of services, the business will ask what's the most cost efficient. IT becomes a broker of services, not necessarily the author of all services. This phase allows business to build the widgets they want off that factory floor [the cloud service]."
There are three requirements for an IT-as-a-service model to be successful, he said. Companies need:
- A petabyte-scalable storage infrastructure that is also compatible with MapReduce technology, for processing huge volumes of data, such as Hadoop;
- A unified analytics platform that functions as a data warehouse and where data scientists can collaborate;
- And a staff of data scientists in each business unit to measure progress and provide analytics information.
"If you've done the first two phases right, building the factory floor [per se], this is the natural transition of how your customers want to go," Mirchandani said.