The architecture of Apache Spark

-

The Run-time architecture of Spark consists of three parts –

1. Spark Driver (Master Process)

The Spark Driver converts the programs into tasks and schedules the tasks for Executors. The Task Scheduler is the part of the Driver and helps to distribute tasks to Executors.

2. Spark Cluster Manager

A cluster manager is the core in Spark that allows launching executors, and sometimes drivers can be launched by it also. Spark Scheduler schedules the actions and jobs in Spark Application in FIFO way on cluster manager itself.

3. Executors (Slave Processes)

Slave processes or Executors are the individual entities on which the individual task of the job runs. They will always run till the lifecycle of a spark Application once they are launched. Failed executors don’t stop the execution of spark job. If any executor failed then that responsibility will be taken up by any other executors.

Other Code concepts:

4. RDD (Resilient Distributed Datasets)

RDD stands for Resilient Distributed Datasets

  • Resilient, i.e. fault-tolerant with the help of RDD lineage graph and so able to recompute missing or damaged partitions due to node failures.
  • Distributed with data residing on multiple nodes in a cluster.
  • Dataset is a collection of partitioned data with primitive values or values of values, e.g. tuples or other objects (that represent records of the data you work with).

An RDD is a distributed collection of immutable datasets on distributed nodes of the cluster. An RDD is partitioned into one or many partitions. RDD is the core of Spark as its distribution among various cluster nodes leverages data locality. To achieve parallelism inside the application, Partitions are the units for it. Repartition or coalesce transformations can help to maintain the number of partitions. Data access is optimized utilizing RDD shuffling. As Spark is close to data, it sends data across various nodes through it and creates required partitions as needed.

5. DAG (Directed Acyclic Graph)

Spark tends to generate an operator graph when we enter our code to the Spark console. When an action is triggered to Spark RDD, Spark submits that graph to the DAGScheduler. It then divides those operator graphs into stages of the task inside the DAGScheduler. Every step may contain jobs based on several partitions of the incoming data. The DAGScheduler pipelines those individual operator graphs together. For Instance, Map operator graphs schedule for a single stage, and these stages pass on to the. Task Scheduler in cluster manager for their execution. This is the task of Work or Executors to execute these tasks on the slave.

6. Distributed processing using partitions efficiently

Increasing the number of Executors on clusters also increases parallelism in processing Spark Job. But for this, one must have adequate information about how that data would be distributed among those executors via partitioning. RDD is helpful for this case with negligible traffic for data shuffling across these executors. One can customize the partitioning for pair RDD (RDD with key-value Pairs). Spark assures that set of keys will always appear together in the same node because there is no explicit control in this case.

Popular posts from this blog

How to change column name in Dataframe and selection of few columns in Dataframe using Pyspark with example

-
Input details: #● File has json records #● Each record has fields: #○ user_id #○ card_num #○ merchant #○ category #○ amount #○ ts ### Below analysis to be done Sample data: +------+--------+---------+--------+----------+-------+|amount|card_num| category|merchant| ts|user_id|+------+--------+---------+--------+----------+-------+| 243| C_108| food| M_102|1579532902| U_104|| 699| C_106|cosmetics| M_103|1581759040| U_103|| 228| C_104| children| M_110|1584161986| U_103| Solution: from pyspark.sql.functions import col # this can be done without using window function creditCardData = spark.read.json(“card_transactions.json”) useridMaxSpendDF = df.groupby(‘user_id’).max(‘amount’) useridMaxSpendDF=useridMaxSpendDF.withColumnRenamed(“max(amount)”,”max_amount”) useridMaxSpendDF=useridMaxSpendDF.withColumnRenamed(“user_id”,”m_user_id”) cond = [creditCardData.user_id == useridMaxSpen...
Read more

What is Garbage collection in Spark and its impact and resolution

-
Image
Garbage Collection Spark runs on the Java Virtual Machine (JVM). Because Spark can store large amounts of data in memory, it has a major reliance on Java’s memory management and garbage collection (GC). Therefore, garbage collection  (GC) can be a major issue that can affect many Spark applications. Common symptoms of excessive GC in Spark are: Slowness of application Executor heartbeat timeout GC overhead limit exceeded error Spark’s memory-centric approach and data-intensive applications make it a more common issue than other Java applications. Thankfully, it’s easy to diagnose if your Spark application is suffering from a GC problem. The Spark UI marks executors in red if they have spent too much time doing GC. Spark executors are spending a significant amount of CPU cycles performing garbage collection. This can be determined by looking at the “Executors” tab in the Spark application UI. Spark will mark an executor in  red  if the executor has spent more than 10% of the time in gar...
Read more

Window function in PySpark with Joins example using 2 Dataframes (inner join)

-
Input details: #● File has json records #● Each record has fields: #○ user_id #○ card_num #○ merchant #○ category #○ amount #○ ts ### Below analysis to be done Sample data: +------+--------+---------+--------+----------+-------+|amount|card_num| category|merchant| ts|user_id|+------+--------+---------+--------+----------+-------+| 243| C_108| food| M_102|1579532902| U_104|| 699| C_106|cosmetics| M_103|1581759040| U_103|| 228| C_104| children| M_110|1584161986| U_103| Solution: from pyspark.sql.functions import col # this can be done without using window function creditCardData = spark.read.json(“card_transactions.json”) useridMaxSpendDF = df.groupby(‘user_id’).max(‘amount’) useridMaxSpendDF=useridMaxSpendDF.withColumnRenamed(“max(amount)”,”max_amount”) useridMaxSpendDF=useridMaxSpendDF.withColumnRenamed(“user_id”,”m_user_id”) cond = [creditCardData.user_id == useridMaxSpen...
Read more