Normalization best practices to achieve improved performance
Normalization helps you clean your data and populate the impact information. The Normalization Engine allows you to run validations and integrity check rules on the CIs and makes them consistent. For example, you can set Normalization rules to specify that all occurrences of Microsoft Word, MS Word, Word should be normalized to MS Word, thus resulting in data consistency.
Ensure the following during normalization:
- You have completed the initial data load.
- The CIs are residing in the respective datasets.
You run Normalization to clean the data, default missing data (if you do not provide a value for data, Normalization assigns a default value to it), and add impact information for every CI.
Other best practices are listed as follows:
Impact is an attribute available as one of the Normalization features, this attribute should be populated. For example, if there is a relationship between Computer System and a software running on it set the impact to 100% when the Computer System goes down the Software goes down. To add impact information see, Creating an impact rule. This impact information helps you to maintain your service model.
The biggest value derived out of the Normalization process is to default the impact information. If you know the impact of CIs, feed it into Normalization.
It is not true that normalization in an inline mode is detrimental to performance. It does take the loader longer to run, but then you do not have a stack of CIs waiting in a batch job to run.
CMDB normalizes CIs as they arrive. If you manually create CIs and have Normalization in inline mode, the slowdown in performance is to a minimum extent. The overall performance does not differ much whether you perform inline normalization or run a batch job for multiple CIs.
If you perform real time normalization, a best practice is to run reconciliation in real time for which you should use the continuous mode. For more information about performing real time reconciliation, see Creating a reconciliation job.
When you perform an initial load of CIs, you should not use inline Normalization mode at dataset level. You should run a batch job. After the initial data load is complete, you can turn to inline Normalization for the selected dataset.