Predictive Functions; Section 6.3, "Predictive Functions - IBM p5 590 System Handbook

Predictive functions

Redundancy in components
Fault recovery
Serviceability features
The p5-590 and p5-595 RAS presents features in all these categories, as
described in the following sections.
6.3 Predictive functions
In a mission-critical application, any outage caused by system failure will have an
impact on users or processes. The extent of the impact depends on the type of
the outage and its duration.
Unexpected system outages are the most critical in a system. The disruption
caused by these outages not only interrupts the system execution, but can
potentially cause data problems of either loss or integrity. Moreover, the recovery
procedures after such outages are generally longer than for planned outages,
because they involve error recovery mechanisms and log analysis.
Planned system outages are also critical in a mission-critical environment.
However, the impact can be minimized by adequately planning the outage time
and the procedures to be executed. Applications are properly shut down, and
users are aware of the service stop. Therefore, there is less exposure when
doing planned outages in a system.
Reliability engineering is the science of understanding, predicting, and
eliminating the sources of system failures. Therefore, the ability to detect
imminent failures, and the ability to plan system maintenance in advance are
fundamental to reducing outages in a system, and to effectively implement a
reliable server.
IBM Eserver p5 590 and 595 System Handbook
142
These are targeted to monitor the system for
possible failures, and take proactive measures
to avoid the failures.
Duplicate components and data paths to
prevent single points of failure.
Provide mechanisms to dynamically recover
from failures, without system outage. This
includes dynamic deallocation of components
and hot-swap and hot-plug capability.
Enable the system to automatically call for
support, and provide tools for rapid
identification of problems.