Data Processing Techniques

In Chapter 2, Artificial Intelligence Storage Requirements, we discussed the layers of a modern data lake and the requirements and possible data storage options for each layer. It became clear that data has to be sent to different data stores to maximize the abilities of AI: building a historical overview and a high-performing queryable source. This means that some work needs to be done with the data before it's suitable for a machine learning model. These data transfers usually happen as ETL steps in a data pipeline. We'll dive into the specifics and possibilities of batch processing in the following paragraphs.

Transactions

In databases, a transaction is a fixed set of instructions that either fail or succeed. Transactions are very useful for data processing since they are reliable and produce no undesirable outcomes. Use them when certain steps are related, or have to be done in a certain order. If a transaction is composed of a hundred...