Providing Open Architecture High Availability Solutions

4.2.2 Compatibility and Interoperability

The definitions of an HA Framework must be clear and unambiguous. They must define the

component interfaces well enough to ensure that different implementations of the same

components will interoperate.

The definitions may leave room for future expandability and growth, but must require standards-

compliant implementations to be compatible. While it is possible that the component boundaries

may change over time, it is important to clearly define the initial models and implementations.

4.2.3 Related Standards Considered

The HA Framework must leverage or recommend existing hardware and software standards and

describe how they are used and integrated into the HA Framework standards. Some may be

considered prerequisites and listed as requirements, while others may emerge as the HA framework

definition evolves. For example:

• Management: WBEM, SNMP, IPMI, etc.

• Configuration/system modeling: CIM, X.731, etc.

• I/O standards: PCI, CompactPCI*, H.100/110, InfiniBand*, etc.

• Networking standards: TCP/IP, ATM, T1/E1, Sonet/SDH, etc.

• Middleware standards: CORBA, etc.

• De-facto standards for operating system, middleware etc.

4.3 System Topologies and Components

Typically the telecom and Internet infrastructure applications are built using clusters of systems

connected by network(s) such as SAN(s) or LAN(s). These clusters contain extra components (e.g.,

redundancy organized in modes such as N+1, 2N, etc.) that can be reconfigured in the event of a

single component failure so that the overall service provided by the cluster is acceptably

maintained.

A system within a cluster may simply be built of standard non-redundant components, or it may

have some internal redundancy to improve its resilience — for example, mirrored discs, redundant

power supply and fans.

The system components can also improve the systems integrity by checking their internal operation

for correctness – for example parity on data paths, and ECC on memory. This kind of checking

improves the speed that faults can be detected and helps contain the damage caused by otherwise

silent faults.