Providing Open Architecture High Availability Solutions

Since the physical repair of a hardware domain will involve direct hands-on interaction between a

system technician and the actual system hardware, having visual guidance for the repair action

directly on the hardware itself is highly desirable. This often takes the form of LEDs and/or other

small display devices that can be controlled through the platform management system.

On-Line Firmware Upgradability

Increasingly, hardware components contain field-programmable devices. This affords an

opportunity to perform repairs and upgrades of the hardware without having to physically remove

and replace components from the system. This can result in significant reductions in MTTRs, as

well as eliminating opportunities for errors being made during a physical hardware swap. Thus,

when programmable devices are used in a system design, having the capability to upgrade the

firmware is highly desirable.

Maintenance of a System Inventory

In any system where hardware components and configuration can change over time, it is critical to

be able to answer the question, “What is currently installed?” This includes a discovery capability

that can detect all installed hardware. Individually replaceable units should be able to report, at a

minimum, what they are (including revision level), what options they contain (if applicable), and

unique tracking numbers. Typically, this information should be made available to the platform

management system.

8.3.6 Event Notification

Throughout all the capabilities of platform management, a common requirement is to provide

notification of significant events to other parts of the system, most notably the management

middleware component of a high availability system. As significant events occur (fault detections,

reconfigurations, etc.), the platform management system should provide information as to what has

occurred. Desirable capabilities of the platform management event notification function include:

• A common format of event messages for all events

• An asynchronous generation of event messages rather than requiring software to poll devices

to learn of events

• A publish/subscribe style interface for communicating events from platform management

system to interested software

• Automatic storage of events in non-volatile memory so they are not lost if there is no current

event listener

• System-wide synchronized time-stamp on events

• Common classification system for events to identify severity, urgency, etc., of event

• Inclusion of enough information with an event to permit immediate fault management action

as well as later analysis for the purpose of root cause analysis on system faults

8.3.7 Additional Useful Hardware Capabilities

The above platform management capabilities have been aimed at the minimum required to support

fault detection, diagnosis, isolation, recovery, and repair of fault domains. Beyond this minimum,

additional platform management capabilities of the hardware in high availability systems can be

provided in order to predict faults and prevent them from occurring in the first place.