6.7.4 Defining Thresholds

About Thresholds


Threshold is a value. You associate the Threshold to a Statistic (PolledData). When data is collected for that Statistic, it is compared with the associated Threshold value. If the collected data value does not suit the Threshold value then it indicates that this kind of data might lead to poor performance of the device or network. Here, the term "suit" is used as you can set up a Threshold value along with a level, such as the maximum value, the minimum value, and equal value.



Threshold Level



Assume you create a Threshold with value 100 and level as Maximum. The collected data being greater than this value indicates a problem.


Assume you create a Threshold with value 100 and level as Minimum. The collected data being lesser than this value indicates a problem.


Assume you create a Threshold with value 100 and level as Equal. The collected data being equal to this value indicates a problem.


These Threshold values act as checkpoints and help in monitoring the collected data.




Threshold Types


Thresholds in WebNMS Performance modules are of three kinds :

Threshold for Long Values


Threshold can be associated with Data identifiers for which the collected data is of type long. Some of such sample Data identifiers are IfAdminStat, IfOperStat, etc in RFC 1213 MIB. The value you provide for this Threshold will be compared as it is with the data collected for the identifier.


Threshold for String Values


Threshold can be associated with Data identifiers for which the collected data is of type string. One of such sample identifiers is SysDescr. You may wish to monitor a change in system description and hence use String Thresholds.


Threshold for Percentage Values


Assume that depending upon the Toner level you want to set the number of pages to be loaded in a printer. When the toner level is 80% of the number of pages to be printed, you may like to be notified.


To achieve this, create a Percentage Threshold with a value 80. Collect data for the Statistics "Toner level" and "Number of pages in the printer". Divide "Toner level" by "Number of pages in the printer" and find percentage. If the resultant value exceeds Threshold value (i.e., 80), then you will receive notification.


Actually, you are associating the second Statistic with the Threshold definition itself. Hence,

    1. Data is collected for first statistic.

    2. When compared with the Threshold value, if the Threshold type is percentage (for the second statistic) data is collected.

    3. Both will be divided and the result will be multiplied by 100, thus giving a percentage.

    4. This resultant value will be compared with 80, i.e., Threshold value.

Note: The Statistics used for percentage calculation must be of same data type. Example: If Statistic one is Counter type then Statistic two should also be of type Counter.



Threshold for Decimal Values


Threshold can be associated with Data identifiers for which the collected data is of type decimal. The decimal value will be got in case of expression OID used in calculations involving divisions. e.g. oid=".$DELTA_TIME"



Threshold for Range Values


Threshold can be associated with Data identifiers over a range of values. We normally specify the upper and lower limits for the range.


    1. in - Threshold event will be raised when the value falls within the selected range . (For eg : if the range for Memory utilization is between 60 to 70, if the value collected is 60, threshold will be violated).
    2. out - Threshold event will be raised when the value falls outside the selected range . (For eg: if the range for Tempature should be between 20 to 25, if the value collected is either 28 or say 18, threshold will be violated and event will be raised).

      Lets consider the example of the monitor the memory utilization of a system. If the memory consumed is between 60 - 70 we need to generate a major alarm. We need to configure the threshold as follows:

<THRESHOLD category="NMSManagement" clrMessage="Memory utilization is normal" kind="range" maxValue="70" message="Memory Utilization is nearing critical" minValue="60" name="Memory-Res-Major" sendClear="true" severity="2" thresholdType="in"

Lets consider the example of monitoring room temperature. If the optimal temperature is 20 - 25 we need to generate a major alarm. we need to configure the threshold as follows

<THRESHOLD category="ATMManagement" clrMessage="Room temperature is normal" kind="range" maxValue="25" message="Room temperature is out of range" minValue="20" name="Memory-Res-Major" sendClear="true" severity="2" thresholdType="out"


Threshold Event Generation


The notification which you receive when the collected value exceeds Threshold value is in the form of a Threshold Event. An event is an occurrence of some action. Hence, whenever Threshold value is exceeded, a Threshold event is generated (handled by WebNMS Fault module).


Also, every Threshold event is associated with a Severity to denote the criticality of the situation. Example: Critical severity could indicate that immediate attention is needed on data collection, etc.


For information on viewing Thresholds, refer to Monitoring Performance Using Thresholds section in User Guide.




Defining Multiple Thresholds


You can associate multiple thresholds with a single Statistic. Doing so helps you in having control over every collected value. For example, you can configure such that if the collected value is above 10, the severity is minor. If collected value is above 20, severity is major and so on. In such a case

  1. Until a threshold reaches its reset value, it will remain in that severity state. As soon as it reaches the reset value, the threshold is reset and waits for the collected data to exceed its limit again.

  2. If another threshold exists of lower severity and if the collected data falls in its limits then that threshold will generate the message.

  3. If a threshold exists with higher severity, then the higher severity threshold will take precedence and its message will be displayed. The threshold with lower severity loses its importance and will not be generated until the threshold with higher severity reaches reset state.



Tips for Creating Threshold


Assume you want to create thresholds for a Router in your network.

  1. Identify the Statistic.

  2. Once you've decided on the Threshold Statistic, establish baseline values. This can be decided by polling using MIB Browser.

  3. After a week, review the data to determine what values are normal for your router.

  4. Use the normal values to determine what the highest acceptable values or thresholds would be.

  5. A good rule of thumb is to set your thresholds 10-20 percent larger than the maximum values.

  6. The threshold values for any Statistic may be applied uniformly across all routers, or they could be customized for groups of routers that have similar characteristics such as core, distribution, and access.

  7. You also need to decide on appropriate notification for threshold violations. Because violations are not considered hard errors, immediate notification is unnecessary. Logging all threshold values and reviewing them daily, usually works well. It is important to investigate repeated threshold violations to determine if a problem can be corrected or if threshold values are too low.



Copyright © 2013, ZOHO Corp. All Rights Reserved.