Providing Open Architecture High Availability Solutions

Typically, these will involve monitoring analog values that can reflect on the health of the hardware

even when a fault has not occurred. For example, a fan may be slowing down, but still functioning

within specifications. This may be indicating a bearing wearing out, and with this warning, the fan

can be replaced before a fault occurs. In another example, monitoring a temperature may indicate

an impending problem before any component has actually failed, triggering a response to bring the

temperature back into a safer range before a fault occurs.

8.4 Open Architecture Solution for Hardware Capabilities

As has been described above, there are a large number of hardware monitoring and control

capabilities required in a fault managed high availability system, but the specific capabilities which

may be required will vary significantly from one system architecture to another.

To support this variety of requirements in an open-standard way, the platform management system

itself needs to have certain key capabilities. Among these are:

• An industry recognized interface to the operating system and management middleware

• The ability to be self-defining. The platform management system should be able to identify

what capabilities it has so that operating systems and management software can adapt to the

specifics of a particular platform.

• Use of an industry recognized management bus for intra-system communication. This allows

for the interoperability of managed hardware components from various vendors with the

overall platform management infrastructure.

• An industry recognized standard for implementing and controlling hot-swap

• An industry recognized hardware interface for key platform components such as power

supplies, cooling units, system boards, and peripheral boards

• An industry recognized interface for platform alarming

• An industry recognized standard for redundant intra-system interconnects

8.5 Standards

8.5.1 IPMI

An example of an open standard, which does a good job of meeting the above hardware capability

requirements is the Intelligent Platform Management Interface (IPMI) specification. This

specification is available at http://developer.intel.com/design/servers/ipmi.

IPMI defines an open-standard abstraction interface and protocol targeted at component level

platform management, which supports the sorts of capabilities described in the preceding

paragraphs.

IPMI also defines a standard interconnect to support communication between platform components

within the hardware layer and between the hardware and operating system layers. Communication

between components is provided by the Intelligent Platform Management Bus (IPMB), which is

typically bused throughout the chassis to connect all platform components including cooling units

and power supplies. The IPMI specification also defines standard communication channels

between the hardware platform and the operating system. These communication channels allow the

operating system, management middleware, and applications to access platform management

capabilities of the hardware. Other key capabilities defined by the IPMI specification include: