Considerations for building private clouds and elements of private clouds in a phased approach

This article focuses on the importance of building private clouds, the constituent elements of private clouds, and the phased construction of private clouds.

Imagine what Amazon Web Services can do in your own private datacenter: Create new disks or servers, or click through the hundreds of templates that support a wide range of operating systems to quickly start a new application instance. This is some of the simple things you need to do to build a private cloud.

A "private cloud" is a market term that refers to mounting an application and server or storage space in a private datacenter and then providing on-demand access to those resources. With this approach, you can quickly deploy new applications, servers, and storage (in the form of logical disks), rather than taking weeks or months to determine requirements, requesting budgets, selecting the right product, and then getting the server before deploying.

What is the importance of building a private cloud?

Migrating your existing application and server infrastructure to a virtual server on a private cloud can save a lot of costs. Today's high-end blade server compared to the previous stand-alone server, due to a significant decrease in energy consumption and has a higher cost performance. You can also take advantage of the data center space (sometimes a small number of blades that can be used in one rack to replace a whole row of servers, possibly halve), deploy more servers in the same space, or reduce data center area and cooling costs.

Constituent elements of a private cloud

The foundation for implementing a private cloud is server virtualization: It is typically running many virtual machines on a single physical server using the Intel x86 architecture and the VMware, Citrix, or Microsoft Super Management programs. Server virtualization can increase CPU and RAM utilization because a large number of virtual machines use the same physical resources, which are originally vacant. For example, a high-speed multi-core server that can extend the memory architecture Cisco USC blades can have up to 32 processing cores and 384GB of RAM, which can run up to 100 virtual machines, each with 4GB memory (this is the number of memory that Windows or Linux servers typically have).

In addition to optimizing resource utilization, Server virtualization enables you to quickly create servers. All you have to do is define the virtual machines and specify the RAM, CPU, and external resources that the virtual server needs. Ideally, you will get a set of predefined templates that you can use to create new virtual machines, eliminating the need to create Windows or Linux servers from the CD/DVD ISO image.

The second constituent element of the private cloud architecture is virtualized storage, or virtual disks are created on demand, and then attached to the desired virtual machines. Virtual disks typically exist as files for large shared file systems; You only need one server virtualization software (VMware vsphere or Microsoft Hyper-V) and a large enough storage array to deploy these virtual storage.

The last, but also important, element is the need for a self-service portal so that users (departments or application teams) can create new virtual machines and disks. For small businesses, the GUI tools provided by virtualization software, such as VMware's vcenter, can implement this functionality. and large enterprises certainly need more advanced software that supports isolation based on user and role access control. The latest release of VMware Vcloud Director or Cisco UCS Manager can meet this requirement.

Building a private cloud in stages

Building a private cloud infrastructure typically consists of the following three phases:

• pilot projects;

• Deploy larger servers and storage infrastructures to migrate existing servers to virtualized environments;

• Implement a true private cloud with self-service portals and implement an on-demand payment mechanism as needed.

These phases may take months or even years to complete, so it is a good idea to take the following measurement steps:

• Select servers and virtualization platforms through market research. I personally recommend using Cisco UCS and VMware. Others may choose to include IBM or HP's rack servers, or Microsoft or Xen Hyper-management technology, depending on the original deployment.

• In most cases you already have a storage array; you just have to make sure it has enough free space. You typically need to allocate several gigabytes of disk space to each virtual machine (in addition to the users who need it);

• After you deploy the new infrastructure and become familiar with it, create a Virtual disk template (for example, typically Windows and Linux servers), and then start using virtual machines to meet the needs of all new servers, simplifying. When people agree that virtualization is a reliable approach, you can start migrating existing applications to virtualized environments to improve performance and reduce datacenter costs.

What else do you need to know?

Server/disk virtualization and private cloud concepts are great tools ... But that's all. You need to be flexible with these technologies, or you'll waste a lot of time and money. Most IT organizations need to change their organizational architecture and the process of consolidating network servers and storage teams because they need to work together to create such an environment.

(Responsible editor: admin)