In this session you will learn the fundamentals of how to apply advanced analytics using Apache spark in Azure databricks. We will focus on how to build and deploy a machine learning model, then I have a look at how you can get started with graph based processing, using graph frames in Apache spark. The combination of big data, machine learning and graph based processing, helps to fully realise the full spectrum of advanced analytics.
You will learn the fundamentals of Spark and how it is enabled on Databricks. Then we will look at how you get started with Machine Learning and Graph based processing
You will be able to begin to work with Machine Learning & Advanced Analytics in Spark.
Spark is one of the most desirable skills on the market. Integrating with big data pipelines is fundamental to the success of Machine Learning with Big Data.
The ability to analyze time series data at scale is critical for the success of finance and IoT applications based on Spark. Flint is Two Sigma's implementation of highly optimized time series operations in Spark. It performs truly parallel and rich analyses on time series data by taking advantage of the natural ordering in time series data to provide locality-based optimizations.
Flint is an open source library for Spark based around the
TimeSeriesRDD, a time series aware data structure, and a collection of time series utility and analysis functions that use
TimeSeriesRDDs can leverage the existing ordering properties of datasets at rest and the fact that almost all data manipulations and analysis over these datasets respect their temporal ordering properties. It differs from other time series efforts in Spark in its ability to efficiently compute across panel data or on large scale high frequency data.
|Spark Version||2.3 and 2.4|
|Python Version||3.5 and above|
Scala artifact is published in maven central:
Python artifact is published in PyPi:
Note you will need both Scala and Python artifact to use Flint with PySpark.
To build from source:
Scala (in top-level dir):
Python (in python subdir):
python setup.py install
pip install .
The entry point into all functionalities for time series analysis in Flint is
TimeSeriesRDD (for Scala) and
TimeSeriesDataFrame (for Python). In high level, a
TimeSeriesRDD contains an
OrderedRDD which could be used to represent a sequence of ordering key-value pairs. A
Long to represent timestamps in nanoseconds since epoch as keys and
InternalRows as values for
OrderedRDD to represent a time series data set.
Applications can create a
TimeSeriesRDD from an existing
RDD, from an
OrderedRDD, from a
DataFrame, or from a single csv file.
As an example, the following creates a
TimeSeriesRDD from a gzipped CSV file with header and specific datetime format.
import com.twosigma.flint.timeseries.CSV val tsRdd = CSV.from( sqlContext, "file://foo/bar/data.csv", header = true, dateFormat = "yyyyMMdd HH:mm:ss.SSS", codec = "gzip", sorted = true )
To create a
TimeSeriesRDD from a
DataFrame, you have to make sure the
DataFrame contains a column named "time" of type
import com.twosigma.flint.timeseries.TimeSeriesRDD import scala.concurrent.duration._ val df = ... // A DataFrame whose rows have been sorted by their timestamps under "time" column val tsRdd = TimeSeriesRDD.fromDF(dataFrame = df)(isSorted = true, timeUnit = MILLISECONDS)
One could also create a
TimeSeriesRDD from a
RDD[Row] or an
OrderedRDD[Long, Row] by providing a schema, e.g.
import com.twosigma.flint.timeseries._ import scala.concurrent.duration._ val rdd = ... // An RDD whose rows have sorted by their timestamps val tsRdd = TimeSeriesRDD.fromRDD( rdd, schema = Schema("time" -> LongType, "price" -> DoubleType) )(isSorted = true, timeUnit = MILLISECONDS )
It is also possible to create a
TimeSeriesRDD from a dataset stored as parquet format file(s). The
TimeSeriesRDD.fromParquet() function provides the option to specify which columns and/or the time range you are interested, e.g.
import com.twosigma.flint.timeseries._ import scala.concurrent.duration._ val tsRdd = TimeSeriesRDD.fromParquet( sqlContext, path = "hdfs://foo/bar/" )(isSorted = true, timeUnit = MILLISECONDS, columns = Seq("time", "id", "price"), // By default, null for all columns begin = "20100101", // By default, null for no boundary at begin end = "20150101" // By default, null for no boundary at end )
A group function is to group rows with nearby (or exactly the same) timestamps.
groupByCycleA function to group rows within a cycle, i.e. rows with exactly the same timestamps. For example,
val priceTSRdd = ... // A TimeSeriesRDD with columns "time" and "price" // time price // ----------- // 1000L 1.0 // 1000L 2.0 // 2000L 3.0 // 2000L 4.0 // 2000L 5.0 val results = priceTSRdd.groupByCycle() // time rows // ------------------------------------------------ // 1000L [[1000L, 1.0], [1000L, 2.0]] // 2000L [[2000L, 3.0], [2000L, 4.0], [2000L, 5.0]]
groupByIntervalA function to group rows whose timestamps fall into an interval. Intervals could be defined by another
TimeSeriesRDD. Its timestamps will be used to defined intervals, i.e. two sequential timestamps define an interval. For example,
val priceTSRdd = ... // A TimeSeriesRDD with columns "time" and "price" // time price // ----------- // 1000L 1.0 // 1500L 2.0 // 2000L 3.0 // 2500L 4.0 val clockTSRdd = ... // A TimeSeriesRDD with only column "time" // time // ----- // 1000L // 2000L // 3000L val results = priceTSRdd.groupByInterval(clockTSRdd) // time rows // ---------------------------------- // 1000L [[1000L, 1.0], [1500L, 2.0]] // 2000L [[2000L, 3.0], [2500L, 4.0]]
addWindowsFor each row, this function adds a new column whose value for a row is a list of rows within its
val priceTSRdd = ... // A TimeSeriesRDD with columns "time" and "price" // time price // ----------- // 1000L 1.0 // 1500L 2.0 // 2000L 3.0 // 2500L 4.0 val result = priceTSRdd.addWindows(Window.pastAbsoluteTime("1000ns")) // time price window_past_1000ns // ------------------------------------------------------ // 1000L 1.0 [[1000L, 1.0]] // 1500L 2.0 [[1000L, 1.0], [1500L, 2.0]] // 2000L 3.0 [[1000L, 1.0], [1500L, 2.0], [2000L, 3.0]] // 2500L 4.0 [[1500L, 2.0], [2000L, 3.0], [2500L, 4.0]]
A temporal join function is a join function defined by a matching criteria over time. A
tolerance in temporal join matching criteria specifies how much it should look past or look futue.
leftJoinA function performs the temporal left-join to the right
TimeSeriesRDD, i.e. left-join using inexact timestamp matches. For each row in the left, append the most recent row from the right at or before the same time. An example to join two
TimeSeriesRDDs is as follows.
val leftTSRdd = ... val rightTSRdd = ... val result = leftTSRdd.leftJoin(rightTSRdd, tolerance = "1day")
futureLeftJoinA function performs the temporal future left-join to the right
TimeSeriesRDD, i.e. left-join using inexact timestamp matches. For each row in the left, appends the closest future row from the right at or after the same time.
val result = leftTSRdd.futureLeftJoin(rightTSRdd, tolerance = "1day")
Summarize functions are the functions to apply summarizer(s) to rows within a certain period, like cycle, interval, windows, etc.
summarizeCyclesA function computes aggregate statistics of rows that are within a cycle, i.e. rows share a timestamp.
val volTSRdd = ... // A TimeSeriesRDD with columns "time", "id", and "volume" // time id volume // ------------ // 1000L 1 100 // 1000L 2 200 // 2000L 1 300 // 2000L 2 400 val result = volTSRdd.summarizeCycles(Summary.sum("volume")) // time volume_sum // ---------------- // 1000L 300 // 2000L 700
Similarly, we could summarize over intervals, windows, or the whole time series data set. See
One could check
timeseries.summarize.summarizer for different kinds of summarizer(s), like
In order to accept your code contributions, please fill out the appropriate Contributor License Agreement in the
cla folder and submit it to email@example.com.
PySpark is an interface for Apache Spark in Python. It not only allows you to write Spark applications using Python APIs, but also provides the PySpark shell for interactively analyzing your data in a distributed environment. PySpark supports most of Spark’s features such as Spark SQL, DataFrame, Streaming, MLlib (Machine Learning) and Spark Core.
1 - Installation & Configuration
This video will show you how to install and configure Apache Spark on a Linux machine.
Notice that in newer versions of Spark, "Slave" has been replaces by "Worker".
2 - Interacting with Apache Spark Cluster
You will learn the general rules on how to interact with an Apache Spark cluster.
3 - Reading in Data
You will learn how to read in different types of data into Apache Spark with pySpark.
4 - Select, Filter, and Sort
You will learn about the most basic operators of pySpark. You will be able to select columns, filter rows and sort your dataframe.
5 - Aggregation
You will learn how to aggregate your dataframe.
6 - Functions
You will learn how to apply functions to your data. With these functions, you can create new or change existing columns in your dataframe
7 - Join, Union, and Pivot
You will learn how to join, union, and pivot your dataframes.
8 - User Defined Functions, Pandas, and Collect
You will learn how to wrangle your data if pySpark has no solution out of the box.
9 - Writing Data
You will learn how to write your Spark dataframes.
#pyspark #python #apachespark
In this video we are installing Debian which we will use as an operating system to run a Hadoop and Apache Spark pseudo cluster.
This video covers creating a Virtual Machine in Windows, Downloading & Installing Debian, and the absolute basics of working with Linux.
2 - Downloading Hadoop
Here we will download Hadoop to our newly configured Virtual Machine. We will extract it and check whether it just works out of the box.
3 - Configuring Hadoop
After downloading and installing Hadoop we are going to configure it. After all configurations are done, we will have a working pseudo cluster for HDFS.
4 - Configuring YARN
After configuring our HDFS, we now want to configure a resource manager (YARN) to manage our pseudo cluster. For this we will adjust quite a few configurations.
You can download my config file via the following link: https://drive.google.com/file/d/11FL12RHSAug_aQtvaG4r2KJP1RhMw3Pk/view
5 - Interacting with HDFS
After making all the configurations we can finally fire up our Hadoop cluster and start interacting with it. We will learn how to interact with HDFS such as listing the content and uploading data to it.
6 - Installing & Configuring Spark
After we are done configuring our HDFS, it is now time to get a good computation engine. For this we will download and configure Apache Spark.
7 - Loading Data into Spark
Having a running Spark pseudo cluster, we now want to load data from HDFS into a Spark data frame
8 - Running SQL Queries in Spark
Let us learn how to run typical SQL queries in Apache Spark. This includes selecting columns, filtering rows, joining tables, and creating new columns from existing ones.
9 - Save Data from Spark to HDFS
In the last video of this series we will save our Spark data frame into a Parquet file on HDFS.
#hadoop #apachespark #bigdata
Learn the method to integrate Kafka with Spark for consuming streaming data amd discover how to unleash your streaming analytics needs
Kafka is a messaging broker system that facilitates the passing of messages between producer and consumer. On the other hand, Spark Structure streaming consumes static and streaming data from various sources (like Kafka, Flume, Twitter, etc.) that can be processed and analyzed using a high-level algorithm for Machine Learning and pushes the result out to an external storage system. The main advantage of structured streaming is to get continuous incrementing of the result as the streaming data continue to arrive.
Kafka has its own stream library and is best for transforming Kafka topic-to-topic whereas Spark streaming can be integrated with almost any type of system. For more detail, you can refer to this blog.
In this blog, I’ll cover an end-to-end integration of Kafka with Spark structured streaming by creating Kafka as a source and Spark structured streaming as a sink.
Let’s create a Maven project and add following dependencies in
<dependency> <groupId>org.apache.spark</groupId> <artifactId>spark-core_2.11</artifactId> <version>2.1.1</version> </dependency> <dependency> <groupId>org.apache.spark</groupId> <artifactId>spark-sql_2.11</artifactId> <version>2.1.1</version> </dependency> <dependency> <groupId>org.apache.kafka</groupId> <artifactId>kafka-clients</artifactId> <version>0.10.2.0</version> </dependency> <dependency> <groupId>org.apache.spark</groupId> <artifactId>spark-streaming-kafka_2.10</artifactId> <version>1.6.3</version> </dependency>
Now, we will be creating a Kafka producer that produces messages and pushes them to the topic. The consumer will be the Spark structured streaming DataFrame.
First, setting the properties for the Kafka producer.
val props = new Properties() props.put("bootstrap.servers", "localhost:9092") props.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer") props.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer")
bootstrap.servers: This contains the full list of servers with hostname and port. The list should be in the form of
host1: port, host2: port , and so on.
key.serializer: Serializer class for the key that implements serializer interface.
value.serializer: Serializer class for the key that implements the serializer interface.
Creating a Kafka producer and sending topic over the stream:
val producer = new KafkaProducer[String,String](props) for(count <- 0 to 10) producer.send(new ProducerRecord[String, String](topic, "title "+count.toString,"data from topic")) println("Message sent successfully") producer.close()
The send is asynchronous, and this method will return immediately once the record has been stored in the buffer of records waiting to be sent. This allows sending many records in parallel without blocking to wait for the response after each one. The result of the send is a
RecordMetadata specifying the partition the record was sent to and the offset it was assigned. After sending the data, close the producer using the
Now, Spark will be a consumer of streams produced by Kafka. For this, we need to create a Spark session.
val spark = SparkSession .builder .appName("sparkConsumer") .config("spark.master", "local") .getOrCreate()
This is getting the topics from Kafka and reading it in Spark stream by subscribing to a particular topic that is to be provided in option. Following is the code to subscribe Kafka topics in Spark stream and read it using
val dataFrame = spark .readStream .format("kafka") .option("kafka.bootstrap.servers", "localhost:9092") .option("subscribe", "mytopic") .load()
Printing the schema of the DataFrame:
The output for the schema includes all the fields related to Kafka metadata.
root |-- key: binary (nullable = true) |-- value: binary (nullable = true) |-- topic: string (nullable = true) |-- partition: integer (nullable = true) |-- offset: long (nullable = true) |-- timestamp: timestamp (nullable = true) |-- timestampType: integer (nullable = true)
Create a dataset from DataFrame by casting the key and value from the topic as a string:
val dataSet: Dataset[(String, String)] =dataFrame.selectExpr("CAST(key AS STRING)", "CAST(value AS STRING)") .as[(String, String)]
Write the data in the dataset to the console and hold the program from exit using the method
val query: StreamingQuery = dataSet.writeStream .outputMode("append") .format("console") .start() query.awaitTermination()
The complete code is on my GitHub.
#bigdata #apachespark #kafka