HP 3000 Manuals HP 3000 Manuals
Configuring the Expert Facility
Many parts of the GlancePlus Expert facility can be customized by editing
the GLANCNFG file. Be careful when making changes because they will
affect the Expert facility's recommendations when you use it. Two main
areas of the Expert facility can be configured: The highlighting of key
metrics on GlancePlus screens, and the performance recommendations made
by the X command.
Highlighting Metrics
You can change the points where the Expert facility highlights metrics on
GlancePlus screens by adding one or more lines to the GLANCNFG file.
These added lines should have the following syntax:
HIGHLIGHT hitem=newvalue
where hitem is one of the items shown in the following table, and
newvalue is the desired value for that item. The hitem field must be
entered exactly as shown--uppercase and underscore characters are
required.
Table 7-2. GlancePlus Highlighting Characteristics
Hitem Default Description
CPU_CM 20 CPU usage in compatibility mode (percentage
of total).
CPU_SW_CM 500 Switches from compatibility mode to native
mode (switches/second).
CPU_SW_NM 500 Switches from native mode to compatibility
mode (switches/second).
CPU_PROC 1
CPU_FOPEN 5 Rate of file opens (opens/second).
CPU_DISPQ 10 Number of processes in the dispatcher queue.
DISC_UTIL 80 Individual disc drive usage (percentage of
total).
DISC_SERV 0.0500 Service time for an individual disc
(seconds).
MEM_CPU 8 CPU usage managing main memory (percentage of
total).
MEM_PFAULTS 30 Memory page faults (faults/second).
MEM_CLOCK 25 Memory clock cycles.
MEM_LIBFAULT 2 Memory page faults on library code
(faults/second).
PROC_CPUSEC 3600 Time of total CPU usage by a process
(seconds).
PROC_CM 80 CPU usage in compatibility mode by a process
(percentage of total).
PROC_SWCM 100 Switches from compatibility mode to native
mode by a process (switches/second).
PROC_SWNM 100 Switches from native mode to compatibility
mode by a process (switches/second).
Table 7-2. GlancePlus Highlighting Characteristics (cont.)
Hitem Default Description
APPL_CPU 20 CPU usage by an application (percentage of
total).
APPL_DISC 50 Disk I/O rate by an application (IOs/second).
APPL_TRANS 120 Transaction rate by an application
(transactions per minute).
Performance Recommendations
The second area in which the Expert facility can be customized consists
of the rules by which Expert facility accesses a system's overall
performance--also known as low level analysis.
The Expert facility defines a set of symptoms wherein each symptom
represents a particular performance situation. A set of rules define the
conditions that increase the probability that a particular symptom is
true. (Computer scientists might recognize the use of fuzzy logic, a
technique in which a condition doesn't have to be clearly true or false,
but might be "a little bit true" or "very much true.") Basically, the
rules consist of comparisons between two items with an accompanying
probability value. If the rule is true, the probability that the symptom
is also true increases by the rule's probability value.
Items are formed either by using constants or by selecting them from a
list of available performance indicators supplied by GlancePlus at each
interval. Items may also be combined into user-defined variables. The
syntax for this section of the GLANCNFG file is described below. For an
example of the CPU bottleneck rules, see below or examine the GLANCNFG
file.
VAR A = TOTAL CPU - BATCH CPU "INTERACTIVE"
SYMPTOM "GLOBAL CPU BOTTLENECK" TYPE=CPU
RULE VAR A > 70 PROBABILITY 25
RULE VAR A > 80 PROBABILITY 25
RULE VAR A > 90 PROBABILITY 25
RULE CPU QUEUE > 3 PROBABILITY 75
To define a user variable, express the following:
VAR [A...Z] = itemid [+,-,*,/] itemid mnemonic
Twenty-six variables can be defined (VAR A through VAR Z). The definition
of one variable may include another item but only if that variable has a
lesser name alphabetically. For example, VAR B can include VAR A, but
not VAR C. The mnemonic is the name used to represent this variable when
the Expert facility prints out reasons to the user.
To define a symptom, express the following:
SYMPTOM symptom name TYPE=symptomtype
where symptomtype can be one of the following: CPU, DISC, MEMORY,
IMPEDE, SWITCH, CM, RESPONSE, or OTHER.
One or more rules defining when the symptom occurs should follow each
symptom. The syntax for defining a rule is:
RULE itemid comparison itemid PROBABILITY percent
where itemid can be a constant, a user variable, or one of the GlancePlus
performance indicator mnemonics.
Comparison is > (greater than)
< (lesser than)
>= (greater than or equal to)
<= (lesser than or equal to)
Percent is a value from -100 to 100.
You may define as many as 50 symptoms and 100 rules at one time.
GlancePlus performance indicators are defined below.
Table 7-3. GlancePlus Performance Indicators
Indicator Description
TOTAL CPU Total CPU usage in the interval (percentage of
available).
SYSTEM CPU CPU usage by system processes (percentage of
total).
SESSION CPU CPU usage by interactive session processes
(percentage of total).
BATCH CPU CPU usage by batch job processes (percentage of
total).
MEM MGR CPU CPU usage to manage main memory (percentage of
total).
DISPATCH CPU CPU usage used in dispatching processes
(percentage of total).
OTHER CPU CPU usage for other reasons (percentage of
total).
CPU PAUSED Time CPU was idle but disc IO occurred
(percentage of total).
IDLE CPU Time CPU was idle and disc IO did not occur
(percentage of total).
PHYS DISC Total physical disc IO rate (IOs/second).
SYSTEM DISC Physical disc IO rate by system processes
(IOs/second).
SESSION DISC Physical disc IO rate by interactive sessions
processes (IOs/second).
BATCH DISC Physical disc IO rate by batch job processes
(IOs/second).
MEM MGR DISC Physical disc IO rate to manage main memory
(IOs/second).
LOGICAL DISC Logical disc IO rate (IOs/second).
DISC UTIL Average usage of all disc drives (percentage).
TRANSACT/MIN Rate of terminal transactions completed
(transactions/minute).
FIRST RESP Average time to first terminal response
(seconds).
RESPONSE Average time to terminal prompt (seconds).
Table 7-3. GlancePlus Performance Indicators (cont.)
Indicator Description
PEAK LOGL IO Logical disc IO rate for busiest disc drive
(IOs/second).
PEAK PHYS IO Physical disc IO rate for busiest disc drive
(IOs/second).
PEAK MEM IO Physical disc IO rate for memory manager on
busiest disc drive (IOs/second).
PEAK UTIL Usage of busiest disc drive on system
(percentage of total).
CPU QUEUE Average number of processes using or waiting for
CPU.
DISC QUEUE Average number of processes waiting on disc I/O.
MEMORY QUEUE Average number of processes waiting on main
memory.
IMPEDE QUEUE Average number of processes waiting on locks,
impedes, etc.
CM CPU Average time spent in compatibility mode
(percentage).
NM CPU Average time spent in native mode (percentage).
CM SWITCHES Rate of switches from native mode to
compatibility mode (switches/second).
NM SWITCHES Rate of switches from compatibility mode to
native mode (switches/second).
TOTAL MIPS Relative processor speed (millions of internal
clock cycles per second).