The Secret Sauce of Cisco UCS: Servers

When I first entered the industry back in 1998, I worked at a regional Internet Service Provider in downtown Seattle, Washington, in the United States.  At that time our “server farm” was nothing more than a set of warehouse shelves with tower-based personal computers connected to the data center network.  When a customer required a new service to be turned up, the company would requisition a new computer, install it, and put it into production.  In this environment, one service was inseparably tied to a single server.  As pictured below, our ISP had e-mail on one dedicated server, web services on another completely separate box, and Domain Name Services (DNS) on others.

Figure 4: Single-Use Servers

 

Rack Mount Servers

If you recall the image of a shelf with computers sitting on it, you could easily come to the conclusion that a great deal of space was wasted.  As the Internet began to grow and mature, the necessity of getting more equipment into the same amount of space began to get a lot of attention.  This led to the rise of rack-mount servers, in which the same (or in most cases better) computing power got packed into compact boxes that fit into a small space in a network rack (usually expressed in RUs or Rack Units, or 1.75 inches).

Figure 5: Rack-Mount and Blade Servers

 

Blade Servers

Continuing the push toward greater computing density, various vendors (such as HP) introduced the concept of blade servers.  In this form factor, a large enclosure containing multiple slots provided centralized power and network connectivity to a series of servers.  These servers, laid out in dense formats called blades, contain the same typical server components, and insert into the enclosure.  Most data center server deployments make use of Rack Mount servers, Blade Servers, or both.

Server Virtualization

Figure 6: Virtualization Vendors

There are few nearly-universal truths of human experience, usually because of the dominance of personalities and long-held opinions, although one such truth would be the following: all resources are finite!  In the computing and networking world, these resources would be space, power, cooling, and of course, budget.   Simply making servers denser is not enough to solve the problem, in part because of server underutilization.  Using the previous example of single-use servers, here is a possible usage scenario:

Figure 7: Server Utilization

In this example, three physical servers, whether rack-mount or blade form factors, are all drawing power, space, and cooling resources, and none are fully utilized.  Virtualization software, using a construct called a hypervisor, can create three virtual machines on a single unit of hardware, reducing the required resources by 66.6%!  The resulting configuration would look roughly like this:

Figure 8: Virtualized Servers

Most industry experts identify VMware as the market leader in virtualization, though other vendors such as Citrix and Microsoft have managed to establish a presence in this fast-growing technology segment.

Related Courses
Cisco Data Center Training

In this article

Join the Conversation

1 comment

  1. Caulker Reply

    A network that is efficiently run will rely on a multiple servers for proper use. A server can be customized to fulfill a certain, specific need. A rackmount server can consist of multiple servers, each with their role to play all placed in one location for better business. If your infrastructure consists of multiple servers, consider a rackmount server and properly stack them for ease and efficient use.