Quality of Service: Predictability
Fujitsu's Predictive Approach
In order to assess the likelihood of a Service Level Agreement (SLA) violation, it is necessary to identify service patterns and results in advance. This enables you to take corrective action before a problem actually occurs, thereby eliminating it or minimising its impact.
An automated approach to Quality of Service Management depends upon knowing the principles of the managed service and the measurements of system behaviour. If you combine this information with a well-defined SLA, a predictive approach becomes possible. Here the gathering, tabulation and knowledge storage of system data is combined with expert reasoning to direct the range of possible options required to match each SLA.
Fujitsu's Quality of Service uses a hierarchy of prediction to manage and answer increasingly complex questions. This in turn provides valuable systems and application information. The following table gives an example of such a hierarchy. The predictive levels are described in terms of the alerts that might be issued, their triggering conditions, and what information is provided to assist in application or system response management.
| Predictive level | Triggering condition | Questions answered |
| Threshold alerts | Measured metric exceeds threshold value | Has an alert threshold been breached? |
| Sequential alerts | Repeated exceeding of threshold value | Is the alert condition persistent? |
| Correlated alerts | Tandem threshold alerts for multiple resources | How pervasive is the alert condition? |
| Pattern alerts | Current trend fits initial segment of previous alert pattern | Is a future alert likely? |
| SLA predictive alerts | Current trend indicates that achieving SLA is in jeopardy | Can SLA target be met? |
| Advisory alerts | (Applies to all alert levels) | What can be done to decrease the likelihood of future alerts? |
