Providing Open Architecture High Availability Solutions

5.3 Dynamic System Model

5.3.1 Introduction

The system model provides the basis for both configuration and fault management and is a critical

component in meeting availability targets within an HA system. The model is typically

implemented in an in-memory database within the management middleware. This complete system

model may use information from models of other components, such as a model of the hardware.

Information about these component models may be made available via OS hooks or hardware

management processors. The discussions on data collection in section Section 5.4 cover this in

more detail.

A system model is an abstract description of component capabilities, interfaces, relevant data

structures, and interactions. The system model dynamically tracks all of the managed components

in the system and incorporates the configuration, state, relationships and dependencies of these

components.

5.3.2 Concepts

The information contained in the system model should contain:

• Actual System Population – The Actual System Population is a reflection of the components

available in the system at any given point in time. This is a dynamic population, especially on

systems with hot-swap and hot-plug capabilities. This is essentially a census of the system

component population.

• Intended System Population – The Intended System Population is a reflection of the

components that have been, are currently, or are expected to be part of the system. It should be

possible to remove components from the intended population if the system has changed to the

point where some of the components no longer could be used in that system.

• Component Information – Contains static information about the component including the

class, product, manufacturer and revision of a component. For physical components such as an

assembly, this information is also known as the field replaceable unit (FRU) information.

• Component State – Contains the dynamic information for individual system components that

are relevant to availability management. For an HA system, the information should be

dynamically updated as the actual state of the components changes. For example, the X.731

state model supports attributes such as administrative state, operational state, usage state(s),

availability status and control status [X.731

• Component Role – In an HA system with redundant components, these components have

roles such as active, standby and spare/unassigned. Active means that the component is

actively engaged in providing service; standby means that the component is idle and waiting to

take over. Unassigned or spare applies to components that are neither active nor standby, but

are available to be made either active or standby. When an active component fails, the standby

component is promoted to take over the active role. The roles are maintained in the system

model.

• Physical Dependencies – System components have physical parent/child dependencies that

need to be understood in order to manage the availability of a service. Children depend on

their parent’s presence and health for them to be assigned work in the system, while the

parents depend on the presence and health of their children for their own proper operation. For

example, a system’s external I/O may depend on an I/O card, which in turn depends on the

system power. Knowledge of these dependencies enables a management system or an operator

to determine how a fault in one system component affects other components within the system.