Method and abstract programming interface for multi-node system discovery, enumeration and configuration
Original Publication Date: 2004-Sep-21
Included in the Prior Art Database: 2004-Sep-21
There exists a family of computer systems such as IBM eServer xSeries EXA-based servers, composed of multiple nodes. Each node is a collection of computing resources (for eg. SMP or CEC) which can be connected together using connectivity ports and cables to form larger systems. These systems can also be attached to remote I/O enclosures (eg. IBM xSeries RXE-100). Hardware partitions can be configured and managed within these systems. Systems management applications (eg. IBM Director) are used to configure, administer and manage hardware platforms. Existing systems management applications have either been designed to manage single-node platforms, or are platform-specific. Systems management of multi-node systems require platform discovery, enumeration and representation of multi-node systems, and configuration of hardware partitions. A generic framework for managing multi-node systems is required that can be integrated into systems management applications. Any method identified to meet these requirements must also provide abstract programming interfaces for use by other systems management tools to operate on multi-node systems. An additional requirement is that the method should support generic multi-node system architectures (for eg. NUMA-based, etc.) Prior to the invention described in this disclosure, there existed no solution to provide a generic framework for managing multi-node systems.
Method and abstract programming interface for multi -node system discovery, enumeration and configuration
1. Invention Summary
Multi-node systems are composed of nodes, and each multi-node system can be configured into hardware partitions. For the purpose of this disclosure a multi-node system will be termed as Complex, and a node will be termed as Node.
The core idea of this invention is to leverage the existing Object Model within the systems management application (eg. IBM Director) framework. The Object Model is extended by creating new Objects to represent multi-node systems, nodes, remote I/O enclosures and partitions. In addition, the existing Event framework within the systems management application is also extended. This invention also includes a platform abstraction component that is used to interface with native mechanisms (provided by systems management applications) to communicate with hardware platforms. This feature allows for support of generic multi-node system architectures.
The extended Object Model and Event framework is used for: discovery and enumeration of multi-node systems and nodes validation of multi-node systems topology definition and configuration of hardware partitions discovery and enumeration of hardware partitions discovery and configuration of remote I/O enclosures generation, filtering and notifications of new Events relating to multi-node systems, nodes and partitions abstract programming interface to manage multi-node systems, nodes and partitions
The invention disclosed here has the following advantages over existing solutions:
Supports generic multi-node system architectures
Leverages existing systems management application framework (Object Model and Events), thereby
reducing the development effort required
Provides abstract programming interface for use by other tools operating within the system management application framework
2. Invention Description
For the purpose of this disclosure, IBM Director will be used as the reference systems management application. IBM Director Managed Objects and Events will serve as the Object Model and Event framework respectively, as used by this invention. The prototype implementation of this invention within IBM Director is composed of Advanced Platform Discovery (APD) and Access Methods. The Advanced Platform Discovery component implements the core idea of this invention. The APD component uses a platform abstraction component to interface with native mechanisms (provided by IBM Director) to communicate with hardware platforms. For the purpose of this disclosure, the platform abstraction component will be termed as Access Methods.
2.1 Discovery, Enumeration and Validation of multi -node systems and nodes
The following describes the functionality and working of the APD code for discovery of multi-node systems (Complexes).
Each discovered Complex and Node will be represented (with a Unique Id) using the following IBM Director Managed Object factory and Man...