|
Following are the steps one can take to ensure that users have consistent and acceptable experience of Data Warehouse platform. This is a high level list.
Establish the Data Warehouse usage windows
Just like source systems are not/partially available for online use, during the end of day batch runs, the Data Warehouse also undergoes batch operations in the form of Extraction, Transformation, Loading, refreshing the views and populating down-stream data marts & systems. While this is happening, the users may not may complete access to the Data Warehouse, as they will get extremely low response time, OR may get wrong data. Therefore, an operational schedule should be published, which provides the 'window of non/partial -use' to the stakeholders.
Typically the windows of non-use will be the time, when Loading is taking place. Users are able to access the Data Warehouse while the Extraction and Transformation process is going on. Similarly the users should be able to access the downstream data marts, when the Loading process in the Data Warehouse is in progress.
Data Warehouse Process components checks
This is generic to any performance management subject. One needs to establish a control panel, which keeps track of the usage/saturation levels of RAM, CPU, and Data Table Spaces etc.
Keeping a proactive check on the Data Warehouse query operations
One can identify the queries (scheduled and ad-hoc), which have taken long time. The same can be shared with the development/user groups. By this sharing, methods can be found to make the queries more efficient OR splitting them OR placing them on lower priority, if the processing load is not justifying the business case. |
