A service level agreement (SLA) policy lets you set standards of performance for your business applications and then monitor deviations from those standards. When deviations occur, an alert is issued and displayed in the Management Console.
Using SLA's you can get information about events like the following:
When the overall fault rate for a transaction is above 1%
When the response time for a service is 15% below its historic norm
When the maximum response time for the consumer of a transaction is greater than one second
You can apply SLA policies to many different objects: transactions, services, endpoints, individual operations, and consumers. You can also apply one SLA policy to any number of objects in your system, for example, to all transactions, or to any subset of the objects mentioned. Not all SLA policies are appropriate to all objects.
When you define an SLA you must specify the following information
It's type: fixed, baseline, or usage.
What objectives the performance of a given object must meet
The period during which the SLA is enforced
The criteria to be used in selecting the objects to which the SLA is applied.
If you need to shut down an endpoint without affecting the state of your SLAs, you can schedule a downtime for the endpoint before shutting it down. Instrument measurements continue to be collected and displayed during scheduled downtimes, but these measurements are not used for SLA evaluations.
When you create an SLA policy, you specify some number of performance objectives, where each objective is based on one of the monitoring instruments. For example, you might specify a performance objective based on the average response time instrument. You also specify a valid range, what would trigger a warning value, and what would trigger a failure value. The Warning Value is optional.
length of the time period evaluated
whether the period is fixed or rolling
when the evaluation occurs
Fixed periods begin and end at defined times and the evaluation occurs only once, at the end of the period. You set the evaluation timing by specifying an exact time for the evaluation to occur, which is also the end of the period. For example, you can set the fixed period to be one month, and start the evaluation on the second day of the month at noon.
Consecutive fixed periods do not overlap.
Rolling periods begin and end relative to the evaluation timing. You set the evaluation timing by specifying an evaluation frequency. At each evaluation, the end of the period is the time of evaluation, and the beginning is the length-of-the-period earlier. For example, you can define a rolling ten minute period with evaluation occurring every two minutes.
Consecutive rolling periods do overlap.