In my last post on the subject of Cloud Computing, I mentioned two ways to slice and dice data in the cloud — depend on query tools to extract data to a local database, or use Data Warehouses to support the transactional system in the cloud. Today, I’ll delve deeper into these two choices for culling meaningful trends and KPIs from data in the cloud.
Whether or not a transactional system is moved to the cloud, the data collected is still necessary for analytical processing. A transaction processing system is optimized to capture the specific transactions as effectively as possible. On the other hand, analytical processing data has to be optimized to allow detection of trends in Key Performance Indicators. Business Intelligence (BI) systems are usually built on the latter. When the transaction system is in-house, an Extract-Transform and Load (ETL) system can be written to automate the transformation of data from highly normalized transactional to denormalized analytical form.