Press "Enter" to skip to content

Category: Hadoop

Flink Table Store 0.3

Jingsong Lee announces a new version of Flink Table Store:

Sometimes users only care about aggregated results. The aggregation merge engine aggregates each value field with the latest data one by one under the same primary key according to the aggregate function.

Each field that is not part of the primary keys must be given an aggregate function, specified by the fields.<field-name>.aggregate-function table property.

Read on for the full changeset.

Leave a Comment

Spark Structured Streaming with Synapse

Ryan Adams builds a demo:

In this post we are going to look at an example of streaming IoT temperature data in Synapse Spark.  I have an IoT device that will stream temperature data from two sensors to IoT hub. We’ll use Synapse Spark to process the data, and finally write the output to persistent storage. Here is what our architecture will look like: 

Click through for the architectural diagram and step-by-step on how to put the demo together.

Leave a Comment

Parallel Loading in Spark Notebooks

Dustin Vannoy answers some questions:

I received many questions on my tutorial Ingest tables in parallel with an Apache Spark notebook using multithreading. In this video and post I address some of the questions that I couldn’t just answer in the YouTube comments. Watch the video for more complete answers but here are quick responses with links to examples where appropriate.

Click through for the video and some text versions. Dustin includes examples for Synapse and Databricks.

Comments closed

Executing Multiple Notebooks in one Spark Pool with Genie

Shalu Ganotra Chadra, et al, explain what Synapse Genie is:

The Genie framework is a metadata driven utility written in Python. It is implemented using threading (ThreadPoolExecutor module) and directed acyclic graph (Networkx library). It consists of a wrapper notebook, that reads metadata of notebooks and executes them within a single Spark session. Each notebook is invoked on a thread with MSSparkutils.run() command based on the available resources in the Spark pool. The dependencies between notebooks are understood and tracked through a directed acyclic graph.

Read on for more information about how you can use it and what the setup process looks like.

Comments closed

Converting Spark RDDs to DataFrames and Datasets

Ashish Chaudhary does a bit of swapping around:

In this blog, we will be talking about Spark RDD, Dataframe, Datasets, and how we can transform RDD into Dataframes and Datasets.

At this point, most of the libraries I know of accept and produce DataFrames. Occasionally you might need to “downshift” to an RDD to work with some specialty library. But in the event you do have one but want to get to another, Ashish has you covered.

Comments closed

Join Types in Spark SQL

Rituraj Khare makes some connections:

In Apache Spark, we can use the following types of joins in SQL:

Inner join: An inner join in Apache Spark is a type of join that returns only the rows that match a given predicate in both tables. To perform an inner join in Spark using Scala, we can use the join method on a DataFrame.

The set of options is the same as you’d see in a relational database: inner, left outer, right outer, full outer, and cross. The examples here are in Scala, though would apply just as easily to PySpark and, of course, writing classic SQL statements.

Comments closed

External Objects in Databricks Unity Catalog

Meagan Longoria adds external tables and views to an Azure Databricks Unity Catalog:

I’ve been busy defining objects in my Unity Catalog metastore to create a secure exploratory environment for analysts and data scientists. I’ve found a lack of examples for doing this in Azure with file types other than delta (maybe you’re reading this in the future and this is no longer a problem, but it was when I wrote this). So I wanted to get some more examples out there in case it helps others.

I’m not storing any data in Databricks – I’m leaving my data in the data lake and using Unity Catalog to put a tabular schema on top of it (hence the use of external tables vs managed tables. In order to reference an ADLS account, you need to define a storage credential and an external location.

Read on for examples of what you can do with this.

Comments closed

Apache Spark Performance Tuning Tips

Amit Kumar shares a few tips with us:

RDD does serialisation and de-serialisation of data whenever it distributes the data across clusters such as during repartition and shuffle, and we all know that serialisation and de-serialisation are very expensive operations in spark.
On the other hand, DataFrame stores the data as binary using off-heap storage, no need for deserialization and serialization of data when it distributes to clusters. We see a big performance improvement in DataFrame over RDD

Click through for several additional tips.

Comments closed

Sharing Results between Notebooks with MSSparkUtils

Liliam Leme provides an answer to a common Synapse Spark pool question:

I’ve been reviewing customer questions centered around “Have I tried using MSSparkUtils to solve the problem?”

One of the questions asked was how to share results between notebooks. Every time you hit “run” in a notebook, it starts a new Spark cluster which means that each notebook would be using different sessions. Making it impossible to share results between executions of notebooks. MSSparkUtils offers a solution to handle this exact scenario. 

Read on to see what MSSparkUtils is and how it helps in this case.

Comments closed

Synapse Runtime for Spark 3.3 Now in Public Preview

Estera Kot has an announcement:

We are excited to announce the preview availability of Apache Spark™ 3.3 on Synapse Analytics. The essential changes include features which come from upgrading Apache Spark to version 3.3.1 and upgrading Delta Lake to version 2.1.0.

Check out the official release notes for Apache Spark 3.3.0 and Apache Spark 3.3.1 for the complete list of fixes and features. In addition, review the migration guidelines between Spark 3.2 and 3.3 to assess potential changes to your applications, jobs and notebooks.

There’s a lot in there, though I did snicker a bit at log4j 2 being more secure than log4j v1 given what we saw last year, though that gaping hole was fixed.

Comments closed