Case Study 7: Server Performance

Implementation Guide › Case Study Examples › Data Modeling Examples › Case Study 7: Server Performance

Case Study 7: Server Performance

This is a typical case study for server performance.

Given the following data source structure:

Indication	Server	Measure	Timestamp
availability	Appserv01	1	03/01/2001 07:44
response time	Appserv01	354.6	03/01/2001 09:58
CPU load	Dbserv02	83%	03/01/2001 12:12
availability	Appserv01	0	03/01/2001 14:26
CPU load	Dbserv02	94.30%	03/01/2001 16:40
capacity	Firewall01	10%	03/01/2001 18:54
response time	Dbserv02	476.89	03/01/2001 21:08
availability	Appserv02	1	03/01/2001 21:24
response time	Appserv01	774.4	03/01/2001 21:48
CPU load	Dbserv01	83%	03/01/2001 21:52

In addition to the above, are the following calculation requirements:

Calculate the % availability of each application server.

The availability of each server should be calculated separately. Therefore, to calculate availability for a single server, it is necessary to receive events for only this specific server. In addition, the data sources contain other performance indicators that are not relevant to availability calculations (response time, capacity, and so on), so the availability indicators as well as the relevant server need to be filtered.

Since the filtering criteria in CA Business Service Insight are Event Type and Resources, translate the filtering criteria from the data source values into a definition of Resource and Event Type.

In this case, the indicator is an ideal value to translate as an Event type in CA Business Service Insight since it describes logically the type of Event. There are a limited number of types, such as, availability, response time, capacity, and CPU load. This means that for the Metrics that calculate availability for the server, the registration is for the availability Event Type.

In this case, when there are a large number of servers and it is necessary to calculate the availability of each server, it follows that each server has to be defined as a Resource. It then needs to be grouped within a Resource group and the Metric will be clustered over that Resource group.

Suggested modeling:

Event name	Availability Event.
Behavior	Reported as change of status of 0 or 1.
Timestamp field	Timestamp (the only time field in the data source).
Resource field	Server (every server that appears in the data source is translated into an CA Business Service Insight resource).
Event type field	Indicator (each of the values in this field are translated into an Event type in CA Business Service Insight. There are four Event types).
Data fields	Measure is 0 or 1 (only for availability records).

The following resource allocations should be defined:

Resource Type attribute	Application Server
Allocation to Contract party	Each application server is allocated to the contract party, where the relevant server runs its application. This enables registration by contract party that retrieves all servers accordingly.
Allocation to Service	Same as above.
Allocation to Resource group	Optional. It is usually necessary to group Resources where clustering is required.

And lastly, based upon all of the above definitions;

Registration by

For the clustered Metric, calculating the availability of each server individually, the registration is by Resource.

In order to be able to meet the above requirement the following criterion is added:

Calculate average response time of application servers for each Contract party separately.

For this requirement, it is necessary to receive Events of response times for all application servers that are part of the group of servers running applications for the specific contract party. Receiving the response time events is done by registering the Event Type created from the indication field with the value "response time". This ensures receiving only events that report on response times of servers.

In order to receive only the Events that report on the relevant servers for a specific contract party, the resources need to be registered through the Contract party allocation.

A resource can be allocated to more than a single contract party, Service or group. Therefore, it may happen that an event that was sent for the calculation of response time as part of the Contract of contract party A, is also a part of the calculations for contract party B.