If you want to refresh the data every hour, you can set up a task to “process” the database by ingesting the data and performing all of the different aggregations against each combination of dimension. Within multidimensional you can choose how often you’ll perform the loading and processing from the underlying data. In doing this you’re developing a “cube”, with multiple different kinds of aggregations. In multidimensional you define different dimensions and then measures that will aggregate along those dimensions. To the user it doesn’t seem like there are two different types. This is because the newer tabular models were designed so that client applications don’t know whether they are querying a multidimensional or tabular model. The differences between the types of data models are mainly found when you’re authoring them. You can also develop tabular data models within Excel and Power BI workbooks that are embedded within the workbook (which is a great way to get started with Tabular). Tabular models can be deployed to SQL Server 2012 to the current version, as well as to Azure Analysis Services. Multidimensional models can be deployed to SQL Server all the way from SQL Server 7 (’98) to the current version. Standard client applications include Excel (via PivotTables, PivotCharts, and cube formulas), Power BI (via live connection and data import modes), and SQL Server Reporting Services. You can interact with these data models through different client applications, including custom developed applications via client libraries.
While they can still do ad-hoc analysis of problems or experiments, they’re best suited for questions you’ll likely be asking again and again (or simple variants of the same question), such as “why were sales so low this month? Why does the company’s gross margin seem to be dropping? Why are my customer deliveries becoming less on time? How do you interact with them? First up, why even create a data model? In essence, they exist to assist you with recurring analysis of your data.
0 kommentar(er)
