Book Excerpt
An excerpt from the case study on
Los Alamos National Labs. LANL
implemented a VMware vCloud based
hybrid Cloud model within their Agency.
Los Alamos National Labs was one of the first government agencies to implement a Hybrid Cloud. Not only were they one of the first government agencies, they were one of the first in the country, period! And the innovator behind it is already implementing it across several Department of Energy Labs. I recently interviewed Anil Karmel, Solutions Architect at Los Alamos and the architect of this project. He is the brains behind one of the first Hybrid Clouds in use today. In fact, he was so ahead of the curve that he actually worked with VMware to improve their vCloud product as it went through the beta testing process before general release to the public. To review, a Hybrid Cloud is a combination of an on-premise private or community Cloud (called a Cloud cell) and a public Cloud (another cell) managed as one large Cloud implementation. I learned from Anil that he did not start out to convert the entire agency. And he didn't set out to build a Hybrid Cloud. He started by virtualizing his then-current infrastructure using VMware. Even at this stage in the evolutionary process, there were some good reasons to move to a Hybrid Cloud (figure 19-1). His first VMware cluster, assembled in 2006, consisted of thirteen HP ProLiant DL585 servers, each with 4 dual-core AMD Opteron processors, 32 GB of memory and redundant network and fiber channel cards. For storage, he used 250 TB of total storage shared across several HP EVA SAN devices. Anil saw the tremendous cost savings in power, cooling and space that were possible by virtualizing an infrastructure. He decommissioned 105 physical servers and shut down 3 data centers just by virtualizing his existing infrastructure. Last, but certainly not least, Anil realized amazing electrical power savings from decommissioning physical servers and datacenters. The actual savings are shown in figure 19-2.
The initial projection called for a return on investment (ROI) of two years, but the actual ROI he achieved was only nine months from the inception of the project! But why start with virtualization? In Anil's opinion, leapfrogging virtualization to go to Cloud only means that you are setting yourself up for failure. Virtualize, then build best practices, understand what that means, then build the Cloud on top of that. When he completed the virtualization project he then moved on to create the Hybrid Cloud environment, using his Private Cloud and the vCloud implementation of the public Cloud provider, Terremark. But what was Anil's motivation? Hadn't he already accomplished a lot with his virtualization cluster? He certainly had! But there were other requirements that came to light: Anil wanted a self-service web portal to automatically request and provision virtual servers. He wanted the green IT savings dynamically computed and displayed on the website. He wanted his cluster to include LifeCycle Management and Chargeback. He wanted to support a variety of operating systems, including Microsoft Windows, Red Hat Enterprise Linux, and Sun Solaris. He also knew that in order to attract customers to his self-service environment, he would need to offer certain value-added services that were not automatically offered in a standard virtualized environment, including:
- Full backup/recovery of virtual machines
- vCenter Management consoled administrators to support the underlying infrastructure
- No ongoing infrastructure refresh costs
- New virtual servers automatically provisioned for the customer
In practice, this meant that once the virtual machines were requested and provisioned from the self-service portal, all the user had to worry about was the security plan for the virtual machine and the system administration and maintenance of the operating system and the applications. The whole point was to use the principle of "Attraction" versus "Authoritarianism". They practiced this principle in the design. Anil said that although they were not fully virtualized, they had designed a system "for the 80%". In other words, while this system didn't cover everyone's needs at LANL, it met the needs of 80% of the users. For them, that was a huge success. Anil told me that there still are many physical servers left at Los Alamos. Many of those users elected to stay on them for one reason or another. Maybe they felt the difficulty of migrating their workload to a virtual environment outweighed the benefits. Or maybe they decided to stay on that physical machine until it was time for a refresh cycle. Anil decided that trying to design a system for 100% of the needs of the engineers at LANL was impossible. Rather than give up, he decided to design a system that was good for 80% of their needs. As you can tell from his savings numbers, even a system designed for 80% of the users can have a huge financial impact on an agency's budget.
