An interpretation type (for example INT, C2RATE, STRING, DATE and so on)
is assigned to a variable based upon the characteristics of the data in that variable. The
interpretation type describes the characteristics of the data in the variable and
determines the default settings for a number of characteristics for the variable, such as
maximum length, format, and so on.
In IT Service Vision supplied tables, the interpretation type is already
assigned to each of the variables. However, if you add a variable to a supplied table or
create a new table of variables, you need to assign interpretation types to your new
variables.
Interpretation
Assigning the appropriate interpretation type to a variable is necessary
not only for proper formatting for reports, but also for proper processing and reduction.
A variable's interpretation type provides or implies the following:
 The default statistics to keep for the variable and how to calculate
them. This is probably the most important implication of interpretation type.
 The default data type of the variable (character or numeric).
 The default maximum length of the variable.
 The default SAS format to use when displaying the data.
IT Service Vision supports two types of tables: INTERVAL and EVENT. The
statistics associated or calculated for an interpretation type depend on whether the
variable is contained in an interval or event table.
An INTERVAL table records information collected over a (usually fixed)
time interval. Thus, interval tables must have a DURATION variable. For example, an
INTERVAL table that records paging subsystem data might include data such as the number of
page faults over the life of the interval (duration) and the amount of paging space used
over the life of the interval.
An EVENT table records information collected for an asynchronous event.
For example, an EVENT table that records task termination events might include data such
as the time of the task termination, the number of page faults over the life of the task,
and the number of I/O operations performed by the task.
The datetime stamp included in an event table records when the event
occurs, while interval metrics contain data sampled (typically) at regular intervals, and
therefore the datetime stamp usually represents the start of the interval.
Interpretation types are described below, along with their default
information, such as type, length, format, and interval and event statistics. Definitions
for the Statistics Codes are provided at the end of the interpretation types. You can also
refer to the Interpretation Types Summary Table to easily browse the types and their
default statistics and format.
AVERAGE 
Arithmetic mean.

Type: Num
Max. Length: 8
Format: BEST12.2
Interval Statistics: Aw
Event Statistics: csA


C2RATE 
Counter to rate. The interpretation type C2RATE is used by the process
step in UNIX to convert a COUNTER to a RATE in an INTERVAL table when the data is
processed into the PDB. For example, if you have recorded odometer miles and the start
time of each reading, the miles is the counter. During the process step, the duration
would be calculated by calculating the difference between the start time of one
observation and the preceeding observation. Then, the miles per second can be calculated
for each observation. This is done by calculating the difference between the odometer
miles of one observation and the preceeding observation and dividing the result by the
duration in seconds.

Type: Num
Max. Length: 8
Format: BEST12.2
Interval Statistics: Aw
Event Statistics: 


COUNT 
Count of events. The COUNT interpretation is an accumulated count of the
number of events, starting at 0, for the interval or event. An everyday example of a COUNT
would be the number of page faults that occur for a job.

Default Type: Num
Max. Length: 8
Default Format: BEST12.
Default Interval Statistics: SA
Default Event Statistics: cSA

Note: When an average is calculated on a COUNT in an interval
table, the resulting value is always a RATE. This is because the only thing that can be
used to normalize the interval observations is the DURATION. Hence, COUNT/DURATION= RATE,
because there is no way to guarantee that the duration of all intervals will always be the
same. 
COUNTER 
Incremental counter across interval boundaries. A COUNTER is an
accumulated count of events, that does not get reset to 0 (i.e. it spans events or
intervals). An everyday example of a COUNTER would be a regular odometer, which only gets
reset to 0 when the available digits are used up. Interpretation type COUNTER has very
limited usage.

Default Type: Num
Max. Length: 8
Default Format: BEST12.
Default Interval Statistics: 
Default Event Statistics: 


DATE 
SAS date value (number of days since 1Jan60).

Default Type: Num
Max. Length: 8
Default Format: DATE8.
Default Interval Statistics: 
Default Event Statistics: 


DATETIME 
SAS date time value (number of seconds since midnight 1Jan60).

Default Type: Num
Max. Length: 8
Default Format: DATETIME19.2
Default Interval Statistics: 
Default Event Statistics: 


DECADDRESS 
DECNet network device address.

Default Type: Char
Max. Length: 200
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


D2RATE 
Delta to rate. The interpretation type D2RATE is used by the process step
on UNIX, to convert a DELTA (a count of something) to a RATE in an INTERVAL table, when
the data is processed into the PDB. A DELTA is a variable that represents the difference
between two monotonically increasing or decreasing values. For example, on a baseball
scoreboard you might have the inning, starting time of the inning, and the runs in each
inning. The "runs" is a delta. The duration would be calculated by finding the
difference between the starting time of one observation and the preceeding observation.
Then the runs per second could be calculated for each observation, by dividing the number
of runs in an interval, by the duration in seconds.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: Aw
Default Event Statistics: 


ENUM 
Enumeration number (number of items within a set).

Default Type: Num
Max. Length: 4
Default Format: 5.
Default Interval Statistics: 
Default Event Statistics: 


FLOAT 
Floating point number.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: 
Default Event Statistics: 


GAUGE 
A GAUGE is a number that represents an instantaneous reading. An example
of a GAUGE would be the number of jobs currently in a given state. An everyday example of
a GAUGE would be the reading on a car's speedometer.

Default Type: Num
Max. Length: 6
Default Format: BEST12.2
Default Interval Statistics: csA
Default Event Statistics: csA

GAUGE differs from INT, in that statistics are gathered for a GAUGE. GAUGE differs from
COUNT in how the average is calculated. In the case of a GAUGE, the average is a simple
average (not divided by duration), because it is an instantaneous reading and does not
represent an entire interval. 
HEXFLAGS 
Binary data flags.

Default Type: Char
Max. Length: 200
Default Format: $HEX.
Default Interval Statistics: 
Default Event Statistics: 


INT 
Integer number. An INT is a number that stands by itself and for which it
makes no sense to perform statistical calculations. An example of an INT might be
something like a hardware model number, the number of CPUs installed, or the
version/release number of a piece of software.

Default Type: Num
Max. Length: 6
Default Format: BEST12.
Default Interval Statistics: 
Default Event Statistics: 


IPADDRESS 
Network device address.

Default Type: Char
Max. Length: 15
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


LABEL 
SAS label for a table or variable.

Default Type: Char
Max. Length: 40
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


MAXIMUM 
Numeric maximum.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: X
Default Event Statistics: X


MINIMUM 
Numeric minimum.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: N
Default Event Statistics: N


NAME 
String containing one word.

Default Type: Char
Max. Length: 64
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


NETADDRESS 
TCP/IP network device address.

Default Type: Char
Max. Length: 11
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


PCTGAUGE 
Gauge that shows a percentage. It is a number between 0 and 1 and it
represents a proportion.

Default Type: Num
Max. Length: 6
Default Format: PERCENT7.2
Default Interval Statistics: csA
Default Event Statistics: csA

Note that the PERCENT, PCTGAUGE, and PCTGAUGE100 are displayed using the PERCENT7.2 SAS
output format. That output format multiplies a number by 100 before displaying it. For
example, a PERCENT variable with a value of .2345 would be displayed as 23.4%. 
PCTGAUGE100 
Percentage gauge that needs to be divided by 100. It is a number between
0 and 1 and it represents a proportion. This interpretation type is used with the process
task on UNIX.Note that the PERCENT, PCTGAUGE, and PCTGAUGE100 are displayed using the
PERCENT7.2 SAS output format. That output format multiplies a number by 100 before
displaying it. For example, a PERCENT variable with a value of .2345 would be displayed as
23.4%.

Default Type: Num
Max. Length: 6
Default Format: PERCENT7.2
Default Interval Statistics: csA
Default Event Statistics: csA


PERCENT 
Numeric value between zero and one. A PERCENT is a number between 0 and 1
and it represents a proportion. An example of PERCENT would be %CPU busy. An everyday
example of a PERCENT could be a fuel gauge.

Default Type: Num
Max. Length: 6
Default Format: PERCENT7.2
Default Interval Statistics: Aw
Default Event Statistics: csA

When choosing between PERCENT and PERCENT100, the distinction simply involves whether
the data coming into the PDB is between 0 and 1 (PERCENT) or between 0 and 100
(PERCENT100).
Note that the PERCENT, PCTGAUGE, and PCTGAUGE100 are displayed using the PERCENT7.2 SAS
output format. That output format multiplies a number by 100 before displaying it. For
example, a PERCENT variable with a value of .2345 would be displayed as 23.4%. 
PERCENT100 
A PERCENT100 is a number between 0 and 100 and it represents the actual
mathematical percentage of a proportion. An example of PERCENT100 would be %CPU busy.

Default Type: Num
Max. Length: 6
Default Format: 5.2
Default Interval Statistics: Aw
Default Event Statistics: csA

When choosing between PERCENT and PERCENT100, the distinction simply involves whether
the data coming into the PDB is between 0 and 1 (PERCENT) or between 0 and 100
(PERCENT100). 
RATE 
Events per second.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: Aw
Default Event Statistics: csA


STRING 
Group of printable characters.

Default Type: Char
Max. Length: 200
Default Format: .
Default Interval Statistics: 
Default Event Statistics: 


SUM 
Numeric sum of the values.

Default Type: Num
Max. Length: 8
Default Format: BEST12.2
Default Interval Statistics: S
Default Event Statistics: S


TIME 
SAS time value (number of seconds since midnight). TIME is treated the
same as COUNT, because TIME represents an elapsed time, which is simply a count of
seconds.

Default Type: Num
Max. Length: 8
Default Format: TIME12.2
Default Interval Statistics: SA
Default Event Statistics: cSA


TIMETICKS 
Count of timeticks in hundredths of a second.

Default Type: Num
Max. Length: 8
Default Format: BEST12.
Default Interval Statistics: SA
Default Event Statistics: cSA


UNIXTIME 
Number of seconds since 1Jan70.

Default Type: Num
Max. Length: 8
Default Format: DATETIME19.2
Default Interval Statistics: 
Default Event Statistics: 


YNFLAG 
A single character, either Y or N.

Default Type: Char
Max. Length: 1
Default Format: $CHAR1.
Default Interval Statistics: 
Default Event Statistics: 


Statistic Codes
The meanings of the statistics codes are as follows:

 No default
<none> Average
C,c Count
D Standard Deviation
h Sum(DURATION*rate**2)
K Variance
M Number missing
N,n Minimum
p Sum(count**2/DURATION)
Q,q Uncorrected Sum of Squares
S,s Sum
V Coefficient of Variance
w Weighted Sum
X,x Maximum
Y Range

Lowercase characters indicate that the statistic is hidden. Hidden
statistics may be required in order to calculate other requested statistics. For example,
if you request the average, the sum and the count must be kept in order to calculate the
average.
As listed above, the w indicates that the variable is weighted.
You can also set alternate weight variables using the Create or Edit Variable Defintiion
window within the window interface, or using the CREATE and UPDATE VARIABLE statements.
Note: The names of variables in the reduction
levels are created by adding these same codes as the final character of the variable name.
For example, if you requested sum and minimum for variable LU62IOC, those statistics would
be named LU62IOCS and LU62IOCN, respectively. Variable names that are shorter than seven
characters are padded with underscores, such that the sum for variable SMFQIO would be
SUMFQIO_S. Blank is used for Average instead of A, such that variable BATCPU would be
BATCPU instead of BATCPU_A. The names are not case sensitive.
In an interval table, statistical calculations for variables with
interpretation type COUNT, TIME, or TIMETICKS are normalized with respect to the DURATION
of the interval. This is because an Average is calculated based on an implied "per
what." For EVENT data this would be per job, per transaction, etc. For INTERVAL data,
the obvious choice of per interval may not be appropriate, since the interval in the data
can vary dynamically for some collectors. Other collectors may record data in fixed
intervals, but the user can change the interval. In these cases the intervals are not the
same, therefore it would be misleading to compare or analyze them as if they were the
same.
By normalizing COUNT, TIME, and TIMETICKS values by the duration of the
interval, we arrive at "per second" values. This results in rates for COUNT
values (for example, page faults per second), and utilizations for TIME and TIMETICKS data
(for example, CPU seconds per second).
If a weighted average is requested for variables with an interpretation
type other than COUNT, TIME, or TIMETICKS in an INTERVAL type table, the statistical
calculations are weighted by the duration of the interval.
If neither of the previous items apply, such as for variables in an
EVENT table, the statistical calculations are not weighted.
Variable
Interpretation Types  Summary Table

InterpretationType Max LengthInterval StatsEvent Stats Format
+++++++
 AVERAGE Num  8  Aw  csA  best12.2 

 C2RATE Num  8  Aw   best12.2 

 COUNT Num  8  SA  cSA  best12. 

 COUNTER Num  8    best12. 

 DATE Num  8    date8. 

 DATETIME Num  8    datetime19.2 

 DECADDRESS Char  6    . 

 D2RATE Num  8  Aw   best12.2 
+++++
 ENUM Num  4    5. 

 FLOAT Num  8    best12.2 

 GAUGE Num  6  csA  csA  best12.2 

 HEXFLAGS Char  200    $hex. 

 INT Num  6    best12. 

 IPADDRESS Char  15    . 

 LABEL Char  40    . 

 MAXIMUM Num  8  X  X  best12.2 

 MINIMUM Num  8  N  N  best12.2 

 NAME Char  64    . 

 NETADDRESS Char  11    . 

 PCTGAUGE Num  6  csA  csA  percent7.2 

 PCTGAUGE100 NUM  6  csA  csA  percent7.2 

 PERCENT Num  6  Aw  csA  percent7.2 

 PERCENT100 Num  6  Aw  csA  5.2 

 RATE Num  8  Aw  csA  best12.2 

 STRING Char  200    . 

 SUM Num  8  S  S  best12.2 

 TIME Num  8  SA  cSA  time12.2 

 TIMETICKS NUM  8  SA  cSA  best12. 

 UNIXTIME Num  8    datetime19.2 

 YNFLAG Char  1    $char1. 
+++++++
