HP 3000 Manuals

HP GlancePlus User's Manual (for MPE/iX Systems)

What Does The Expert Facility Do? 

The data displayed on each GlancePlus screen is examined by the Expert
facility, and any indicators that exceed the normal range for the size of
system are highlighted.  Since the highlighting feature adds negligible
overhead, it is permanently enabled.

A global system analysis is performed based on data obtained from a
single sample.  This can be a response to an on-demand request (you
pressed the X key), or might occur automatically following each screen
update, if the Expert facility is in continuous mode.  During global
analysis, all pertinent systemwide performance indicators are passed
through a set of rules.  These rules were developed by top performance
specialists working on the HP 3000.  The rules were further refined
through use on a variety of systems of all sizes and configurations.  The
response to these rules establishes the degree of probability that any
particular performance situation (called a symptom) could be true.

If the analysis is performed on demand, any symptom that has a high
enough probability of being true is listed along with the reasons (rules)
why it is probably the case, as in the following example:

     XPERT Status: 75% CHANCE OF GLOBAL CPU BOTTLENECK
     Reason:       INTERACTIVE > 90.00 (96.4)

This says that "most experts would agree that the system is experiencing
a problem when interactive users consume more than 90% of the CPU."
Currently, interactive use is 96.4%.  Since the probability is only 75%
(not 100%), some additional situations are not true.  (In this case, the
number of processes currently starved for the CPU might not be high
enough to declare a real emergency.)

Summarization of expert analysis is available when the Expert facility is
used in continuous mode.  When this is enabled, global analysis is
performed after each screen update, but messages are not given unless 
there is a problem.  When one or more problems are detected, a
single-line summary is added to the current display screen.  In
generating this one-line summary, each symptom is counted as one or more 
types of problems (CPU, DISC, MEMORY, LOCKS, Compatibility Mode, Switch
Rate, or Response Time).

The probabilities for all symptoms of a given type are added to arrive at
the state of the resource.  If the state of the resource is low enough,
the resource is considered normal, if higher, it is evaluated as
MODERATE, HIGH, or VERY HIGH. The continuous-mode Expert facility display
is a single line showing any problem areas and their states.  For
example,

     XPERT: |CPU IS VERY HIGH| |SWITCHES ARE HIGH|

In addition to generating one-line displays, the Expert facility also
retains a summary of how often each type of problem occurred in each
state.  By typing O (for Overall Summary), you can generate a summary
report showing the totals since the Expert facility was put into
continuous mode.


          Figure 6-1.  Xpert Overview Screen 

High level analysis can be performed only if the Expert facility is
enabled for high level--use the V command:  XLEVEL=HIGH. After a global
analysis in which a problem type was not normal, the processes that
executed during the last interval are examined.  If an action can be
suggested that might improve the situation, the action is listed as
follows:

     XPERT:  Status 75% CHANCE OF GLOBAL CPU BOTTLENECK.
     Reason: INTERACTIVE > 90.00  (96.4)
     Action: QZAP pin 122 (PASXL) for MEL.EELKEMA from "C" to "D" queue.

Actions will not be instituted automatically since you may or may not
agree with the suggestions.

If the Expert facility is enabled for high level analysis, you can
request additional information when you request an expert diagnosis using
the X command.  Following the initial display, you can type X again to
display the process data analysis.  This analysis often includes the
process or processes that are probably causing the problems as well as
those that are suffering from the problems.  For example,

Enter the X command:

     XPERT:  Status 75% CHANCE OF GLOBAL CPU BOTTLENECK.
     Reason: INTERACTIVE > 90.00  (96.4)
     Action: QZAP pin 122 (PASXL) for MEL.EELKEMA from "C" to "D" queue.

Press RETURN to continue, then press X for more details, or O for an
overview summary.  Press X again:


            

Even if no action is recommended, it can be helpful to display the high
level analysis.  The analysis might indicate why no action should be
taken at this time.  You might discover that even though your system CPU
is 100% busy due to interactive users and a CPU bottleneck is indicated,
no action is recommended since no process is starving for the CPU.
Continued use of the expert high level analysis can teach you some of the
subtleties of managing performance.  You may not always agree with the
suggested actions, but the information can help you understand your
system better.

There is one level of analysis available beyond the high level discussed
above.  If after viewing the high level analysis, you still want more
information, you can respond by pressing X again.  This will give you the
"Dump Everything" level analysis.

Usually, high level analysis is only done on areas of system performance
that the Expert facility determined to be a problem.  Processes will not
be examined for disc problems unless there was an overall disc problem
indicated.  In the "Dump Everything" level analysis, every possible type
of problem is examined to determine even a remote possibility that a
process is causing a problem.  This can be useful if you are interested
in potential database-locking problems on your system, but the problems
are not affecting enough of the system to be considered serious.

MPE/iX 5.0 Documentation