Building Private Cloud Novice Guide

Third, building the internal cloud because resources are shared so that some companies that are not listed are also involved

Stanley thinks the way to build a private cloud has nothing to do with technology. is inextricably linked to the organization-when marketing does not want to share the server with human resources, the Finance Department is unwilling to share information with any department. In this situation, it is difficult to combine them, and the utilization of the cloud is very inefficient. In the face of such challenges, it director Marcos Athanasoulis has come up with an innovative way to adapt participants to the idea of sharing resources on the Linux cloud infrastructure, combining information well, making them individual groups, He thinks this method will help the hardware as well.

At Harvard Medical School, Athanasoulis, the 1000 chief Information officer invited, said it faced some unique challenges. Telling a lab what technology is lacking in authority is a limitation, but it can be done if a lab wants to configure their own infrastructure. So when Harvard Medical School walked into the cloud concept four years ago, it wanted to build this pattern in a shared way. Let their research results attract labs that want to build their own high-performance computing or cloud environments, if we can't provide a viable alternative.

With this approach, if the lab buys 100 nodes on the cloud, you can guarantee that you can use those capabilities. But if these capabilities are idle, other workloads can run on them.

"I told them-you have the hardware, but if you let us integrate him into the cloud, we will manage the device and upgrade and patch it for you," he said. This proved to be a good selling point.

To support this approach to hardware, Athanasoulis says, Harvard Medical School uses platform computing's platform LSF workload automation software. "This tool gives us the ability to create queries and suspend tasks that run on dedicated hardware nodes so that the owners of the hardware can be assured that the paused work will be stored again."

Start deployment after understanding the service

If resources are not shared, the cloud is inefficient and meaningless if the service is not a priority. IBM, for example, assessed the different types of workloads and risks, benefits, and the cost of migrating workloads to different cloud models before embarking on a possible cloud plan, said Fausto Bernardini, director of IT strategy and architecture at IBM's cloud computing Services Division.

Tom Bittman, vice president and renowned analyst at Gartner Consulting, argues that many companies often think about building a private cloud from a product perspective before considering service and service requirements-a complete reversal of the process.

If you really want to build a private cloud, Bittman says, you have to understand what your service is, what the service level agreement is, and the cost and cloud roadmap must be linked together. This actually involves your perception of whether the service is evolving towards the cloud computing type.

Common services with relatively static interfaces (even if your enterprise relies heavily on these services) are those that you should consider as cloud computing, private or public cloud, and email is an example, Bittman.

"I could use a lot of these services, but they have nothing to do with the work inside my company." This is the kind of service that can be independent-I don't want her to be close to the company. I want to keep it as independent as possible, easy to use, and can be applied from the self-service interface. If I can customize this service over time, I will do my best to standardize the service, "Bittman said.

On the contrary, services that are closely related to businesses and that continue to focus on innovation initiatives are not the first choice for cloud services. Bittman that "the goals of these services are privacy and integrity." They can't migrate to the clouds. They can use lower-level cloud functions, such as raw calculations, but the company's interface does not point to cloud mode.

Once you know what services are suitable for the cloud, and how much time the private cloud will take you to understand, you are ready to build the enterprise model, and you can build a private cloud from a technical perspective.

Last step: Service Management and usage management
Finally, Gartner defines four classes of components to build a private cloud.

At the bottom end is a resource layer of infrastructure, platform, or software. Virtualization will soon come to mind, but if you are consolidating resources into resource pools, virtual machines are not the only option. For example, the rapid reconfiguration technology is another option.

Above the resource pool is the resource management layer, Bittman said, "This is how we manage the pool of resources in an automated way," Bittman also stressed that for VMware environments, this step is about how to use the VMware Distributed resource Scheduling table.

"These two levels are quite mature," Bittman said. You can find such products in the marketplace, although the current field of resource management is less competitive.

The next step is the service management layer. "This is a more complex area of demand," Bittman said. I need tools that can help me with service monitoring, and I need tools that can help me transform resource pools into service levels. Finally, I must be able to bring the user a service-level interface, which can display performance or usability, so that the service management to achieve this.

When you think about how to build a private cloud, it's important to understand the gap between demand and product availability, "Bittman stressed." "For example, VMware is doing a great job of helping users manage virtual resource pools, but they don't know anything about services." VMware's Vcenter Appspeed is the initial attempt to start dabbling in service management.

"What we really need is good service regulation, but it hasn't been achieved yet," Bittman said.

At the top of the hierarchy is the use of management, which is closely related to the User Self-Service service interface. "It provides a list of services that allows users to manage their subscribers." This interface must somehow be bundled with cost and cost returns, or at least tied to the calculation-this step is closely related to the service management layer.

Cost returns are a very painful challenge for private cloud providers, but they do not have a long-term disregard. "From a technical standpoint, it's like a joke--what is the basis of the charge?" said Bittman. But from a political and cultural standpoint, frankly, if I wanted to migrate to cloud computing, I would suggest migrating to the cost return model.

Finally, it's about the business model.
While cloud providers must consider flexibility, automation, self-service services and cost returns, they should not be too rigid in the current cloud development phase. "We're going to see a lot of businesses that are engaged in pure clouds and pure bittman, and many are in the middle," he said. All these entanglements at one point ' what's the benefit? '"。

For example, the Wentworth-douglass hospital in New Hampshire State Dover, USA, is using Vblock systems from Cisco, EMC and VMware to build their private cloud, But their focus is mainly on server consolidation rather than on the self-service configuration of the software, says Scott Havena, a hospital's network operations manager.

"Maybe we'll end up turning to software as a service, and we're doing a lot of work on automation, but it's slow to implement within the enterprise because the cloud model is so advanced that it takes a lot of time for the entire enterprise to understand its essentials," he said.

As Harvard Medical School says, "The reason people use our cloud is because it provides attractive value, which is not bad for it."

"Edit Recommendation"

Teach you how to build a private cloud (upper) cloud computing model and the choice of cloud architecture cloud computing architecture and logic from the perspective of virtualization

"Responsible editor: Wang Yong TEL: (010) 68476606"