Multi-Protocol Approach to Systems Management Functional Test

Publishing Venue

IBM

Abstract

Current systems management service processors often provide numerous interfaces and protocols to support the inspection and control of the operation of a managed system. The lowest-level protocols provide hardware layer interfaces to the managed device, where the higher-level interfaces and protocols may provide management capabilities to many multiple systems and devices. The concurrent use of these multiple interfaces and protocols for functional verification of systems management firmware/hardware provides several benefits.

Multi-Protocol Approach to Systems Management Functional Test

The concurrent use of these multiple interfaces and protocols for functional verification of systems management firmware/hardware provides several benefits:
- deterministic validation of data presented by interfaces from low-level providers
- methods to determine model instances from mapping of low-level data providers
- validation of state specific data based on interdependency of interfaces
- efficiency of testing common functional blocks throughout interface hierarchy

Using this multi-protocol approach to functional test can provide a more consistent, deterministic and complete functional test of a systems management device.

The model for a systems management device is constructed from static information about the managed system (SMBIOS, SDR, etc) and the current state of the system and device, allowing validation of the model data by comparisons to the source providers after appropriate transformation.

Specifications exists mapping the source provider data to data model instances and values presented by other interfaces, and mapping CIM (Common Information Model) model data to other legacy protocols. Model data is collected from low-level data providers (IPMI, SMBIOS), consolidated by a central provider (CIMOM) and redistributed through legacy interfaces (SLP, Dot, SNMP, CLI, WebUI) and CIM protocols (SM-CLP, CIM-XML, WS-Man).

Greater efficiency is realized in testing function common among multiple interfaces collectively (or randomly) through the available interfaces. Functional tests as a sequence of atomic operations may be indifferent to the protocol used allowing random or systematic selection of an interface that supports the operation, therefore increasing test coverage within a time constraint.

Functional testing of Systems Management protocols can be limited by the test procedures ability to predict the current state of the systems management model for the system under test. The data model for systems management is constructed from static information available about the system being modeled (i.e. VPD/FRU data, SDR data, and the actual configuration of the system's hardware), as well as the current state of the system under test at the time the test is being performed. The data model is distributed upward from low-level data providers (e.g. IPMI), potentially through Mid-layer data providers (i.e. dot, SNMP, CIMOM), to high-level protocol providers (i.e. CIM-XML, SMASH-CLP, WS-Man). Through-out the process of distributing the system management data model errors could be introduced by systems management firmware errors in transforming, translating, or by introduced propagation delay (stale caches) into higher level data model representation. Producing a accurate representation of even the static data model for comparison for a given system configuration would be difficult. Given the difficulty in producing static data mode...