Some months ago bithw1 posted an interesting question on my Github about multiple SparkSessions sharing the same SparkContext. If you have similar interrogations, feel free to ask - maybe it will give a birth to more detailed post adding some more value to the community. This post, at least, tries to do so by answering the question.
Some time ago on my Github bithw1 pointed out an interesting behavior of Hive integration on Apache Spark SQL. To not delve too much into details now, I can tell that the behavior was about not respected DataFrame schema. Our quick exchange ended up with an explanation but it also encouraged me to go much more into details to understand the hows and whys.
Apache Spark SQL provides advanced analytics features that we can find in more classical OLAP-based workloads. Below I'll explain one of them.
Most of RDBMS are able to store JSON documents in columns of JSON-like type. One of them is PostgreSQL that can keep JSONs in one of 2 columns (JSON or JSONB) and that natively enables querying of JSON document attributes. As we'll see below, with a little bit of effort we can implement similar behavior in Apache Spark SQL.
Nested data structure is very useful in data denormalization for Big Data needs. It avoids joins that we could use for several related and fully normalized datasets. But processing such data structures is not always simple. Fortunately Apache Spark SQL provides different utility functions helping to work with them.
One of previous posts in SQL category presented window functions that can be used to compute values per grouped rows. These analytics functions are also available in Apache Spark SQL.
Some recent posts covered important Spark SQL options for RDBMS: partitioning and write modes. However they're not the only ones available for this data storage.
Some months ago I presented save modes in Spark SQL. However, this post was limited to their use in files. I was quite surprised to observe some specific behavior of them for RDBMS sinks. Especially for SaveMode.Overwrite.
Some weeks ago I presented correlated scalar subqueries in the example of PostgreSQL. However they can also be found in the Big Data processing systems, as for instance BigQuery or Apache Spark SQL.
In programming a simple is often the synonymous of understandable and maintainable. However it doesn't always mean efficient. One of examples of this thesis is nested loop join that is also present in Apache Spark SQL.
Prior to Spark 2.2.0 release, the data processing was based on a set of heuristic rules ignoring the typology of the data. But the most recent release brought a tool well known from the RDBMS world that is a Cost-Based Optimizer.
DataFrame can either be loaded and saved. And Spark SQL provides, as for a lot other points, different strategies to deal with data persistence.
It's time to continue the exploration of operator optimizations of logic plans in Spark SQL. After the first part describing optimizations from A to L, this post covers remaining letters.
Pushdown predicate is one of the most popular optimizations in Spark SQL. But it's not the single one and their main list is defined in org.apache.spark.sql.catalyst.optimizer.Optimizer abstract class.
User Defined Functions are not the single way to extend Spark SQL. The second solution is offered by User Defined Aggregate Functions.
Spark SQL has a lot of "hidden" features making it an efficient processing tool. One of them are statistics.
The optimizer in Spark SQL helps to improve the performance of processing pipelines. One of its techniques is predicate pushdown.
Chain of responsibility design pattern is one of my favorite's alternatives to avoid too many nested calls. Some days ago I was wondering if it could be used instead of nested calls of multiple UDFs applied in column level in Spark SQL. And the response was affirmative.
User Defined Types (UDT) described in one of previous posts aren't the single customization possibility in Apache Spark SQL. The other possibility are User Defined Functions (UDF).
Spark SQL reading from RDBMS is based on classic JDBC drivers. Thus it supports some of their options, as fetchsize described in sections below.
You want to learn data engineering but have no idea where and how to start in this wide domain? Check if Become a Data Engineer can help you 💪