More Content - Including Podcasts

Wednesday, June 8, 2011

HP Discover and the Mayo Clinic's Journey Into The Cloud

Presented by Paul Steger, Section Head of Enterprise Computing Services at the Mayo Clinic.

Discussion prefaced by Paul noting that they are not at the destination yet, but underway on the journey. Legal department insisted Paul read the slide disclaiming his presentation before he begins.

Mayo clinic operates in three states, widely distributed across the continent.
Invests over $769M per year in education and research.

Why did they care about virtualization? They found in 2004 their average utilization of server infrastructure was less than 10%, with Windows server, being 82% of all servers at less than 5.1%.

Their datacentres were out of space and capacity, virtualization became the default answer, and VMWare was chosen as the vendor of record to assist in consolidating systems hardware.

Worked initially within the refresh cycle to be conservative in approach.

They've just completed a blitz cycle, and removed over 420 physical servers reclaiming 167kWh in the data centers, they are now 82% virtualized. They found an astonishing. 27% of servers removed really weren't needed anyhow. This was an awakening for the infrastructure teams, and having this made aware to the rest of IT and IS is vital.

As of January 2008, any new server requests were by default virtual, unless a business and technology case can be strongly made to the contrary.

support clause with vendors placed in contract language clarifying that any new services or products WILL be supported by the vendors in a virtual environment.

ROI calculations include power savings, hardware cost avoidance, software cost avoidance. When asked about the future of cloud computing, and when management claims "we have a cloud" then answer is "no, not yet. But were heading there." the question is raised, what is a cloud, and the answer proposed is that it is still just servers, storage, and connectivity. Quoted the NIST definition of cloud computing which I've referenced in a previous blog posting.

Mayo is only looking at private cloud computing, and as IaaS and Platform as a Service as the options. When working with the other groups and customers, they have to agree to lease end dates, there will be no perpetual leases within their cloud. Important business model for governance of the infrastructure to ensure that their is capacity to expand as needed.

Mayo Infrastructure team cannot charge back for their services, but Cloud technologies offer metering capabilities so that Paul can state "here is what your service is costing Mayo Clinic; do you still need this?"

Service delivery improvement will eventually happen with a focus on automation, process, and best practices governance. The goal is to have an SLA for server provisioning of a gold image replica in 24 hours. They know they can come in under the SLA, but are managing expectations.

Dev/Test is the environment where they started because of the lower daily use and decreased customer risk. It gave a platform to help measure the ROI as well. Research and Education groups have been great candidates as well because they work load is seasonal so the elasticity of cloud computing is a perfect fit.

They plan to have multiple business models of cloud computing to be able to make business sense to the different groups they support, different compliance and business operations requirements.

Process of developing the roadmap was to draft a cloud computing white paper, developed a reference architecture in a one page picture, documented simply the requirements, and was granted agreement to complete a POC. Once completed, they need to white paper that case to evaluate the business case and ROI. Leveraging the existing infrastructure as much as possible was key to the architecture and successful acceptance.

Service catalogue clearly identifies the goldmimage available, that's all available in the cloud. The secret sauce is the orchestration layer in the private cloud. this is a plan though, because they do not yethave a service catalogue.

The initial requirements were defined in use cases primarily around "push button" provisioning, and the governance around this provisioning.

Mayo sat down with HP, Cisco, and IBM; HP won the opportunity to create the POC, and it was completed in April 2011. Testing and training will continue through August 2011. They will be working with two internal teams at Mayo to execute to POC testing. Noted that HP came to the table as a real partner in this, and was realistic around the costs, understanding how much of an opportunity this POC is for both Mayo and HP.

Paul has gone on a "speaking tour" to educate different units about what this POC is, what Cloud means to Mayo, where they are planning tom go, and how people can get on board. Subsequently, there will be another roadshow to live demo what they have accomplished.

Reiteration of the point that clients are not being given permanent systems. Question had come up if systems could be archived for later use if not needed for a long period. The answer given to clients is no, when the lease is up, if you can't renew with a business case, it goes away and the resources are reallocated to another business service.

Next steps for Mayo are to re evaluate skill sets, staffing, and general infrastructure support. Many questions and learning is needed before they can broadly expand the offerings.

Access and security
Change control and regulatory compliance
High availability
Software license management
Dynamic provisioning and deprovisioning to meet demand bursts
Organizational structures, teams, and job skills
Cultural acceptance of such a change

Paul posed the question, are we ready for this change in our organization?

- Posted using BlogPress from my iPad

Location:Las Vegas, NV

No comments: