International Business Machines Corporation

Wade C Allen

Paul Joseph Reder

Jeremy P Goodwin

Douglas Toltzman

Robert Louis Nielsen

Antonio Perrone

Mark Clarence Zelek

Rosalind Radcliffe

Paolo Deidda

David Anthony Fresquez

James L Panian

For monitoring a data processing system, a sequence of critical conditions with increasing severity indexes is defined. Each critical condition is detected when a corresponding state parameter exceeds a threshold value with a predefined persistence over time. The persistence may be specified by a minimum number of repeated occurrences of this event, with a maximum number of consecutive non-occurrences that are allowed between them. Whenever critical conditions are detected, the monitoring application retrieves the severity index of the last detected critical condition with the highest severity index, and calculates an incremental index indicative of how closely the state parameter is approaching the next critical condition. An indicator of the health of the system may be generated by subtracting the severity index of the last critical condition plus the incremental index from an optimal value.

Monitoring method, system, and computer program based on severity and persistence of problems

In an Open System Interconnection (OSI) environment, an agent Common Management Information Protocol (CMIP) platform maintains an up to date tree of all of the Managed Object Instances (MOIs), in order to perform scoping functions. The agent platform receives a message from an OSI manager which includes a base MOI and levels below the base MOI to receive the message. The agent platform recurses through the tree below the base MOI to determine which MOIs should receive the message. If the message is delete, the children of MOIs to be deleted are also deleted. The children of undeletable MOIs are not deleted. A current, accurate tree is maintained by the agent platform, which can register a new MOI in the tree when a valid request is received from the MOI or the manager. The agent platform can also deregister an MOI upon valid request from the agent associated with the platform.

Object independent scoping in an open system interconnection system

Information on GDMO managed objects, including class hierarchy and the name binding templates, is compiled into a naming table. This naming table is built from an inheritance table and the name binding template information. The inheritance table indicates the inheritance hierarchy of the managed object classes. The compiler builds the inheritance table as a square matrix with a row and column for each class, and the classes arranged in an ordered list from highest to lowest in hierarchy. The compiler enters a one at each position in the inheritance table to indicate that the managed object class in the row inherits from the managed class in the column. With the inheritance table as a ready reference, the compiler builds the naming table from the name binding templates available as listed in the GDMO source files. The naming table is a square matrix identical to the inheritance table except the entries in the naming table are not ones and zeros. Instead, entries in the naming table are lists of name binding templates that may be used to instantiate a managed object instance in the row class under a managed object instance in the column class.

Compilation of information contained in GDMO name bindings

Problems with prior OSI (Open Systems Interconnected) event forwarding discriminators (EFD) are solved in accordance with this invention by creating a new type of EFD, called a notification forwarding discriminator (NFD). An NFD works just like an EFD in that it receives notifications emitted by managed objects and determines what types of notifications should be forwarded. However, the forwarding mechanism in an NFD is completely different. An NFD has two operations that a management application can use to control subscription to notifications, 'start' and 'stop'. An NFD forwards notifications via responses to a solicited 'start' operation without signaling the end of the 'start' operation. Hence, a 'start' operation will always be pending completion. The manager and agent track this pending status of operations between a manager and an NFD. Both the manager and agent maintain a pending completion list for each link between a manager and agent. There is a one-to-one correspondence between each manager and the link over which requests go to a given NFD through an agent. Multiple applications using the manager may send multiple start operation requests through the manager, through the agent and to the NFD.

Notification forwarding discriminator

Managed Object Instances (MOIs) in a network of OSI computing systems are created, or deleted, without disrupting their interaction with other managed object instances. This is accomplished by a multi-phase commit process operation where in the first phase, a create or delete request is verified, and the CMIP platform places the created or deleted MOI on a pending list in the CMIP platform, along with information identifying the location of the MOI in a hierarchy of MOIs. In the second phase, the object being created or deleted is asked to accept, or reject, the request. The object responds saying that it accepts or rejects the request. If the request is acceptable, the CMIP platform finishes the object's connection to the tree hierarchy, or disconnects the object from the tree. Then the CMIP platform sends back a message to the object to acknowledge the creation of the object, or to tell the object to delete itself. Finally, the CMIP platform sends a request completion message to the manager.

Multi-phase commit processing for creation and deletion of managed

A new event forwarding discriminator (EFD) EFD named the persistent/impervious event forwarding discriminator (PI EFD) is described herein. The PI EFD is monitored by the agent infrastructure in such a way that if the agent goes down, and thus the PI EFD goes down, the agent rebuilds or restores the PI EFD with all its attributes. The PI EFD immediately creates an event notification indicating a create PI EFD event has occurred. This event notification is of the type passed by the PI EFD to the manager. The manager then knows the poll its other EFDs at the agent and recreate them as necessary. To foil rogue managers, the PI EFD can not be deleted by other managers. Also, the PI EFD has limited attributes that can be changed. A rogue manager can only add himself to the destination list; it can not otherwise change the attributes of the PI EFD.

Growing community of inventors

Durham, NC, United States of America

Wade C Allen