need more memory or CPU, however, you have to
physically add new memory or a new processor, or you
can transfer the application to a different server.
Figure 13-8 Bare-Metal Application Stack
While bare-metal application deployment is an effective
way to isolate applications and get consistent
performance characteristics, it’s very inefficient. In fact,
many of these traditional server deployment models
result in enormous amounts of waste in regard to power
and cooling of servers when they are not under load. It’s
not uncommon to find a server during non-peak hours
using less than 10% of its capacity, which is not very cost-
effective.
Some applications, however, really need a dedicated
server. Big data applications benefit greatly from a
dedicated server environment; most Hadoop clusters
consist of numerous one-rack-unit servers all running in
parallel. These workloads consume 100% of the
resources of the server, making them poor candidates for
virtualization.
VIRTUALIZED APPLICATIONS
Virtualization was created to address the problems of
traditional bare-metal server deployments where the