Skip to main content

The Power of POWER Virtualization

The rationale for the growing attention being paid by IT to system virtualization is strongly manifested in the emergence of hot-swap server hardware. Starting with storage drives, the popular notion of improving system availability through hot-swap hardware now extends to PCI expansion cards and memory modules. The common attribute for all of these hardware resources is dynamic configurability. In the minds of most IT managers, this is best represented by the ability to replace a failing I/O adapter while the system continues to run. For these IT managers, high reliability, availability, and serviceability (RAS) are the dominant issues.

Virtualization provides an easy way to extend the notion of dynamic reconfiguration of system hardware without the need to reboot systems. With resource virtualization, dynamic reconfiguration can be extended to include adding or removing processors. As IT workloads change, operations managers are theoretically free to add or remove processors, memory, and I/O adapters to and from logical partitions (LPARs) of computer systems. Making such abstract notions real and actionable is what the new IBM® Systems Director family of virtualization solutions is intended to do. The Systems Director family is explicitly intended to consolidate the management of virtualized resources across a broad spectrum of technologies represented by IBM System x™, BladeCenter®, IBM System p™, System i™ and System z™ servers.

For the Intel- and AMD-based servers and blades that fall under the System x and BladeCenter product line offerings, virtualization is a relatively new technology that, for the most part, is now being pursued through the trapping and emulation of privileged Operating System (OS) instructions for full virtualization. In such an environment, any OS can run in an LPAR without any changes. This fits with the typical X86-usage scenario of running a standard shrink-wrapped OS distribution. Nonetheless, the process of trapping and emulating OS instructions introduces significant overhead in context-switching and parameter-checking operations.

POWER Partitioning
On the other hand, robust partitioning and virtualization have been strong components of PowerPC Architecture for quite some time. Besides being a significantly more mature technology on POWER5™, POWER5+™, and PowerPC 970MP microprocessors, virtualization follows a very different approach from the notion of trap-and-emulate.

In line with datacenter usage scenarios, virtualization on POWER requires running a specialized version of an OS that is tailored for scalability and performance, rather than a generic shrink-wrapped version of the OS. In particular, virtualization is implemented through a technique dubbed paravirtualization. For an OS to run in an LPAR, it must be augmented with the ability to make calls that invoke the Hypervisor.

The POWER Hypervisor is always active and running as core System p firmware. It implements and controls an abstraction layer between the physical hardware resources and an LPAR. Micro-partitioning—at the level of a tenth of a processor—and the Virtual I/O server for shared virtual I/O devices are added through the Advanced Power Virtualization (APV) package.

“POWER hardware and firmware support adding and removing virtual processors, entitled processor capacity, uncapped processor weight, memory, and I/O devices for logical partitions while the operating systems are running.”

Hypervisor Convergence
Numerous advances have recently bolstered the industry-leading virtualization capabilities of POWER machines. POWER hardware and firmware support adding and removing virtual processors, entitled processor capacity, uncapped processor weight, memory, and I/O devices for logical partitions while the operating systems are running. A major feature of System p and System i machines is now a common Hypervisor that provides common functionality. Nonetheless, each brand continues to retain a unique value proposition.

The key to leveraging all of the new operational management features, which are now part of the Advanced POWER Virtualization scheme for IBM System p servers, is the integration of System Director through the IBM Virtualization Manager Extension with either the Hardware Management Console (HMC) or the Integrated Virtualization Manager (IVM). The HMC is a set of firmware tools that manage platform resources via messages to the hypervisor and the operating systems of the partitions. The Integrated Virtualization Manager operates in a similar fashion to the HMC, but it does not require the use of a Hardware Management Console and manages a single system.

The HMC creates the initial configuration definition, controls boot and termination of the various partitions, and provides virtual console support. Since the moving of resources involves cooperation between the HMC and the operating system, that makes the HMC the control point for dynamic reconfiguration. What’s more, this dynamic reconfiguration will grow more extensive over the coming year.

Also growing more extensive will be the notion and usefulness of an LPAR. Currently, an LPAR is a virtual container for a logical machine and represents a basic bare-metal system. Nonetheless, the notion of a virtual machine is only part of what the IBM Systems Director family of systems management tools can handle. Beyond the basic virtual machine is the workload running on that machine.

An important component of the Systems Director family is the IBM Enterprise Workload Manager (EWLM). EWLM is designed to dynamically monitor and manage distributed workloads in heterogeneous environments based on well-defined business goals. By utilizing that notion and extending the idea of an LPAR from a container of a bare metal virtual machine to the container of a fully provisioned operational workload partition, IBM plans to leverage the advanced RAS features of POWER Architecture, and leading edge virtualization technologies more deeply with respect to critical corporate initiatives, such as Service Oriented Architecture (SOA) and thereby provide IT with a solid roadmap for success.


We're here to help

Easy ways to get the answers you need.

  • Contact me

Or call us at: 1-866-883-8901
Priority code: 6N8AG62W

Content navigation

Related links