• No results found

Calculation Method

columns. So that column provides the one value per resource. (You can specify the mean or the weighted mean calculation method, but not both. They are mutually exclusive.)

Note: The underlying tables at the resource level store both the weighted mean and the unweighted mean for the response time, throughput, and custom

objectives. (At the higher nodes, the only mean that is kept is the one that you specified as the calculation method.) 4

For availability, the ETL job writes one row for each resource, and the row contains the uptime for the resource (in seconds) divided by the expected duration, expressed as a percent. So the rows in the availability table look like this:

RESOURCE

DOMAIN MACHINE UPTIME

US Operations HAL6000 (299/300*100=) 99.67 US Operations Server/01 (298/300*100=) 99.33 US Operations ALLSERV (290/300*100=) 96.67

The calculation method (that you specify next) obtains its input from the UPTIME column, and that column provides the one value per resource.

Note: The example shows how original intervals shorter than five minutes are consolidated into five-minute intervals. Similarly, original intervals longer than five minutes are prorated into five-minute intervals.

Also, regardless of the interval length and even if the original intervals are five minutes), the ETL job redistributes the data so that the new intervals are on the following five-minute boundaries: :00, :05, :10, and so on. 4

Calculation Method

For an objective, the calculation method obtains the appropriate values from the objective’s underlying tables (Availability, Response Time, Throughput, and Custom Objective), performs the appropriate calculation, and assigns the result to the objective’s indicator in the five-minute row in the component’s main table in the contract database.

For more information about the tables, see “The Contract Database” on page 237.

You can choose one of these calculation methods: maximum, minimum, mean, weighted mean, sum, percentile, redundant, or serial.

Understanding Contract Objectives 4 Calculation Method 29

3 Maximum or Minimum:

For the availability objective, the maximum or minimum calculation method obtains the values in the UPTIME column (99.67, 99.33, and 96.67), takes the maximum (99.67) or minimum (96.67) of them, and assigns that value to the availability indicator.

For the response time, throughput, or custom objective, the maximum or minimum calculation method obtains the values in the MAX column (27, 23, and 36) or MIN column (23, 21, and 31), takes the maximum (36) or minimum (21) of them, and assigns that value to the response time indicator, throughput indicator, or custom objective indicator.

3 Mean:

For the availability objective, the mean calculation method obtains the values in the UPTIME column (99.67, 99.33, and 96.67), takes the mean of them (98.56), and assigns that value to the availability indicator (one value for each five-minute interval).

The response time, throughput, or custom objective calculates the mean of the child node or resource indicators. It then assigns the resulting value to the response time indicator, throughput indicator, or custom objective indicator (one value for each five-minute interval).

3 Weighted Mean:

The response time, throughput, or custom objective calculates the weighted mean of whatever the child node or resource indicators are, using the weights for the respective child node or resources, according to the following formula:

x = P

n

i=1

w

i

1 x

i

P

n i=1

w

i

where

x

is the weighted mean, the numerator is the weighted sum, and the denominator is the sum of weights.

For example, the calculation of the weighted mean (TI) for the throughput objective’s indicators is as follows:

TI = [(26.9*106) + (22.9*106) + (31.3*108)] / 106 + 106+ 108 TI = 2851.4 + 2427.4 + 3380.4 / 320

TI = 8659.2 / 320 TI = 27.06

It then assigns the resulting value to the response time indicator, throughput indicator, or custom objective indicator (one value for each five-minute interval).

Note: When an objective requests the calculation of the weighted mean, the child nodes are not required to calculate their indicators as weighted means. The parent node calculates both the weighted sum and the sum of the weights in order to determine the weighted mean for the objective.4

3 Sum:

For the availability objective, the sum calculation method obtains the values in the UPTIME column (99.67, 99.33, and 96.67), takes the sum of them (295.67), and assigns that value to the availability indicator (one value for each five-minute interval).

30 Calculation Method 4 Chapter 3

For the response time, throughput, or custom objective, the sum calculation method obtains the values in the SUM column (125, 110, 169), takes the sum of them (404), and assigns that value to the response time indicator, throughput indicator, or custom objective indicator (one value for each five-minute interval).

3 Percentile:

For the availability objective, the percentile calculation method obtains the values in the UPTIME column (99.67, 99.33, and 96.67), sorts them in ascending order (96.67, 99.33, 99.67), identifies or calculates a value that corresponds to the percentile that you specified (in the GUI), and assigns that value (99.33) to the availability indicator.

For the response time, throughput, or custom objective, the percentile

calculation method obtains the values in the MEAN column (25.0, 22.0, 33.8), sorts them in ascending order (22.0, 25.0, 33.8), identifies or calculates a value that corresponds to the percentile that you specified in the GUI (say, 50th), and assigns that value (25.0) to the response time indicator, throughput indicator, or custom objective indicator (one value for each five-minute interval.

This method is most useful if you have a large number of inputs, ideally more than 100. In this example, because 3 is such a small number of input values, you would probably not select the percentile calculation method.

Note: The percentile calculation uses the same default algorithm that is used by PROC UNIVARIATE (specifying option PCTLDEF5).4

For more information about the percentile calculation method, see “Percentile”

on page 217.

3 Redundant:

If the objective type is availability and if the relation of the component’s resources is redundant, it is a good idea to select the redundant calculation method. (A redundant relation is one in which all of the resources are in a pool and serve as a backup to one another; thus, the component is available even if only one of its resources is “up.”)

For the availability objective, the redundant calculation method obtains the value in the UPTIME column (99.67, 99.33, and 96.67), applies the redundant

“formula,” and assigns the resulting value (100.00) to the availability indicator.

For the response time objective, throughput objective, or custom objective, the redundant calculation method is not offered as a choice.

3 Serial:

If the objective type is availability and if the relation of the component’s resources is serial, it is a good idea to select the serial calculation method. (A serial relation is one in which each of the resources is both unique and required;

thus, the component is available only if every one of its resources is “up.”)

For the availability objective, the serial calculation method obtains the value in the UPTIME column (99.67, 99.33, and 96.67), applies the serial “formula,” and assigns the resulting value (95.71) to the availability indicator.

For the response time objective, throughput objective, and the custom objective, the serial calculation method is not offered as a choice.

3 Custom Code:

You define custom code that allows user-defined SAS code to be written in order to assign a value to the objective’s indicator.

Note: For more information about the Maximum, Mean, Minimum, Sum, and Weighted Mean methods, and the formulas used to calculate them, see “Maximum, Mean, Minimum, and Sum” on page 217.

Understanding Contract Objectives 4 Understanding More about Objectives in Higher-Level Nodes 31

For more information about the redundant calculation method, see “Redundant” on page 217.

For more information about the serial calculation method, see “Serial” on page 218.

For more information about the Custom Code method, see “Custom Code for the Calculation Method” on page 221.4

Note: If the five–minute interval is in an outage period, then the indicator’s value is not calculated. Instead it is assigned, based on the type of the outage (either scheduled, unscheduled, or excused). For more information about outages, see “What is an

Outage?” on page 231. 4

For more information about tables in the contract database, see “The Contract Database” on page 237.