The virtual datacenter architect must be able to take a holistic approach to datacenter design. This means that, for every decision made, the architect must understand how the environment, as a whole, will be impacted.
An architect is required to be, at the very least, familiar with all aspects of the datacenter. They must understand how the different components of a datacenter, such as storage, networking, computing, security, and management, are interconnected, as shown in the following diagram:
The holistic approach to datacenter design
It has become very important to understand how any decision or change will impact the rest of the design. Identifying dependencies becomes an important part of the design process. If a change is made to the network, how are computing, management, and storage resources affected? What other dependencies will this introduce in the design? Failing to take a holistic approach to design can result in unnecessary complications during the design process and potentially costly fixes after the design is implemented.
You have been engaged to design a virtualization solution for a financial organization. The solution you are proposing is using 10 GB Converged Network Adapters (CNA) to provide connectivity to the organization's network in three 1U rackmount servers. The organization needs to separate a Virtual Local Area Network (VLAN) that is currently configured to be delivered over the CNA onto a physically separate network to satisfy a new compliance requirement. A 1 GB network will provide sufficient bandwidth for this network, and the network should be highly available. Single points of failure should be minimized.
To support this compliance requirement, you, the architect, must take a holistic approach to the design by answering a number of questions about each design decision. Some questions are as follows:
Are there available network ports in the current rackmount servers, or will a network card need to be added? If a card must be added, are there available Peripheral Component Interconnect (PCI) slots?
Will a dual port network card provide sufficient redundancy, or will the network need to be separated across physical cards? Are there onboard network ports available that can be used with a PCI network card to provide in-box redundancy?
Has the physically separate switch's hardware been obtained? If not, how long before the equipment is received and deployed? Will this have an impact on the implementation schedule?
How will the virtual switch need to be configured to provide the connectivity and redundancy required?
The impact can be fairly significant, depending on some of the answers. For example, let's say that the 1U rackmount server will not support the required network adapters needed to satisfy the requirement, and a different 2U rackmount server must be used. This then raises more questions, such as: "Is there sufficient space in the rackmount to support the new server footprint?"
What if the requirement had been that the applications connected to this network be virtualized on separate physical server hardware and storage? What parts of the design would have to change? The architect must be able to understand the dependencies of each part of the design and how a change in one place may affect other areas of the design.
As you think through these questions, you should be able to see how a change to a requirement can have a deep impact on many other areas of the design. It becomes very important to identify requirements early on in the design process.
