Virtualization Management - Oracle Solaris 10 System Virtualization Essentials - page 310

CPUs for most of the quarter. When the end of the quarter nears, two more CPUs
can be assigned to the VE.
This flexible assignment of resources becomes even more valuable when the un-
expected occurs. Continuing with the previous example, a surprisingly successful
fiscal quarter might put even more demand on that VE. Upon noticing that four
CPUs are not sufficient to handle the workload, a fifth CPU can quickly be as-
signed to the VE, thereby enabling the VE to complete its task on time. Business-
critical workloads will benefit the most from this ability to quickly respond to
changes in resource consumption.
Configurable resource controls are necessary to achieve operational flexibility.
In many virtualization implementations, these controls may be modified while
the VE and its application are running. The resource controls commonly available
today were described in Chapter 1, “Introduction to Virtualization,” and include
exclusive CPU assignment or CPU capacity limits, RAM and virtual memory as-
signment or limits, process schedulers, and others.
Some virtualization technologies have had resource controls added to them over
time. Others have inherited existing, proven technologies. For example, Oracle
Solaris Containers take advantage of the resource controls that have been part
of Solaris for many years, but also use new controls implemented more recently.
Some of these tools can use other existing features, such as time-based com-
mand execution, to make scheduled changes to resource controls. Still other tools
can be used to make changes if certain conditions exist. For example, a CPU might
be added to a VE's set of CPUs if CPU utilization exceeds a certain threshold. An
example of this approach was described in Chapter 6, “Oracle Solaris Containers.”
Alternatively, if the utilization of all CPUs becomes too high, the virtualization
management system might move a VE to a different, under-utilized system. The
latter scenario is described later in this chapter, and is part of the topic of workload
mobility, which was described in general terms in Chapter 1. Workload mobility
relies on virtualization because it takes advantage of the weaker bond between a
VE and the hardware on which it runs.
Conversely, data center flexibility can be improved if tools can be used to free up
resources that are no longer needed. Those resources can then be used to handle
new or growing workloads. In some cases, a workload in a VE might become less
utilized over time. If it no longer needs all of the RAM assigned to it, the virtualiza-
tion tools should make it possible to dynamically change the memory assignment
or memory cap for that VE without a service outage. This tactic might make room
for a new workload, which may be deployed as another VE on the same computer.
In contrast, without the flexibility achieved by virtualization, the purchase of a
new computer might be necessary to handle the new workload.
Resource controls can certainly improve operational flexibility, but other work-
load properties can help, too. Physical location can be considered a property of