Flipping around traditional ETL (Extract-Transform-Load) on its head is not a new practice. ELT (Extract-Load-Transform), where processing is handled in the database, instead of the ETL server, has been proven to enhance performance in many types of data warehousing deployments.
For example, Oi, a leading telecom provider in Brazil, implemented an enterprise data warehouse (EDW) consolidating information on 36 million customers, speeding response time to customer requests. The right-time EDW also enabled Oi to rapidly launch a successful new service offering, which made it easier for customers to recharge their pre-paid accounts for telecom service.
By implementing ELT with Informatica’s pushdown optimization capabilities for this Teradata data warehouse, Oi accelerated its data warehousing loading process two-fold. This has led to even more timely updates of Oi’s customer information, while lowering costs.
ELT, or pushdown optimization, plays an important role in turbocharging performance by reducing the amount of data being moved around unnecessarily, and it reduces costs by taking advantage of existing computing resources in the database or data warehouse. However, it certainly isn’t optimal for all situations. Where the data lies, the characteristics of the data itself, and how it needs to be transformed dictate whether ELT or ETL will provide the best performance for any given situation. Even within a single data warehousing deployment, in most cases ELT methods need to be combined with ETL to ensure optimal overall performance. Hence, it’s important to have a data integration platform that can support both modes easily. More on this in my next posting.