Introduction to Integrity Virtual Machines

Increased utilization and scalability
Very few data centers achieve utilization rates above 40 percent. This is usually the result of two
common practices: First, there is typically a one-to-one relationship between applications and
servers. Second, many enterprise-class applications are resource-intensive, but the high demand is
only for short periods. Virtual machines can be used to improve the ratio of applications to physical
resources since it can automatically allocate those resources to the right application at the right time.
Many businesses have applications that are cyclic in nature, such as payroll and end-of-month billing.
Frequently, the cycles for the applications do not coincide. Servers are normally configured to handle
the peak loads, causing lower average utilization of the server. Integrity VM automatically allocates
larger shares of the system resources to applications during the peak usage times. Similarly, during
off-peak times, applications can automatically receive a smaller allocation of system resources. In this
way, physical resources are utilized by those virtual machines where applications are experiencing a
high demand for those resources.
With Integrity VM, a virtual machine is allocated a specified amount of memory, one or more virtual
CPUs, and a set of virtual I/O devices; hence, multiple, isolated virtual servers share the underlying
physical resources. Integrity VM optimizes this sharing by using HP’s proven workload management
technology, automatically allocating resources to virtual machines based on demand and entitlement
guarantees set for that VM.
Many existing virtualization technologies can provide multiple, single-CPU virtual machines, with one
or more virtual machines per physical CPU. Such virtualization enables much more flexibility in the
data center in the form of faster provisioning and higher resource utilization. However, this
paradigm, depicted in Figure 2, is limited in what class of virtual server it can provide. Because this
approach does not scale to more than one CPU per virtual machine, it is not practical for enterprise-
class and mission-critical applications that require the scalability inherent in symmetric multi-processor
servers (SMPs).
Figure 2. Single-CPU virtual machines