large amount of processor and network resources because those resources can be
shared. So while one OS is not using the resource, others can. When it comes to
resources that are divided up, such as memory and storage, it’s possible to add them to
and remove them from a running virtual machine as needed. This type of elasticity is
not possible in traditional physical deployments, and with virtualization hosts
generally architected to have far more resources than in a physical OS deployment, the
scalability, or maximum resource that can be assigned to a virtualized OS is much
larger.
The consolidation of operating system instances onto a smaller number of more
powerful servers exposes additional virtualization benefits. With a reduced number of
servers that are more powerful but more highly utilized, organizations see reduced
datacenter space requirements, which leads to energy savings and ultimately cost
savings.
Many organizations have long struggled with a nontechnical aspect of their
datacenters, and that is licensing. I cover licensing in detail later in this chapter, but
when you have thousands of individual servers, each running a single operating
system, it can be hard to track all of the licenses and hard to know exactly what
version you need based on the capabilities required. Most important, it costs a lot of
money. With virtualization, there are ways to license the virtualization hosts
themselves and allow an unlimited number of virtual machines, making licensing of
the OS and management software far more cost-effective.
Another challenge with a single operating system per physical server is all the islands
of resources that you have to manage. Every server has its own local storage, and
somehow you have to protect all of that data. Utilizing centralized storage such as a
SAN for every physical server is possible but typically cost prohibitive. It’s not
practical to purchase Fibre Channel HBAs (cards that enable connectivity to Fibre
Channel switches), Fibre Channel switches to accommodate all of the servers, and all
of the cabling. Take those same servers and reduce the number of physical servers
even tenfold using virtualization, and suddenly connecting everything to centralized
storage is far more realistic and cost effective. The same applies to regular networking.
Implementing 10Gbps networking in a datacenter for 100 servers is far more possible
than it is for one with 1,000 servers.
On the opposite side of the scale from consolidation and centralization is the
challenge of isolating workloads. Consider a branch location that for cost purposes has
only a single server to host services for the local workers. Because there is only a
single server, all roles have to run on a single OS instance without virtualization,
which can lead to many complications in configuration and supportability. With
virtualization, that same server can host numerous virtual machines, with each
workload running in its own virtual machine, such as a virtual machine running a
domain controller and DNS, another running file services, and another running a line-
of-business (LOB) service. This allows services to be deployed and isolated to standard
best practices. Additionally, many remote offices will deploy two virtualization servers