Spark Interview Questions for Freshers 2021

Spark is a program that analyses massive quantities of data and distributes the results across many computers. It can execute tasks a hundred times faster and includes more than 80 high-level operators, making it simple to develop parallel apps using Hadoop, Apache Mesos, or Kubernetes stand-alone systems or in the cloud.

Apache Spark is a fast-paced technology that powers many of today’s most high profile services. It offers scalable processing power and excellent optimisation capabilities, which are vital in an interview situation where you need to show your future employer how well prepared for the job they can expect from their newest member.

In this blog, we’ll cover all of our top 25 picks for getting those tricky questions right – so read carefully before heading into the interview room with people who will decide whether or not YOU get hired.


  1. Tell us, what is Apache Spark?

Apache Spark is a fast, easy-to-use framework engine that offers big data processing and analysis. It also has modules for graph processing and machine learning capabilities, all built into the system directly, so you don’t have to worry about installing another program. It also has modules to process graphs, machine learning, streaming SQL, among other things, which make it a powerful tool to use across many different industries with ease.


  1. What are the Apache Spark Features?

Speed: Spark is known for being extremely fast, but it truly shines in the area of processing speed. Apache Spark reduces read-write operations by almost 100x while performing calculations on disk and can perform ten times faster than other systems when using memory instead of as your primary data storage medium.

Dynamic: Spark provides 80 high-level operators, which help in the easy development of parallel applications. With these tools, you can create and distribute content with ease across all platforms, including mobile.

In-Memory Capability: The Spark data processing engine’s in-memory capabilities allow it to execute faster, cache previously loaded queries and reduce the time required for disk I/O.

Reusability: Spark codes can be reused for batch-processing, data streaming and running ad hoc queries. 

Lazy Evaluation: Spark transformations done using lazy evaluation in the RDDs are a great way to increase system efficiency. This is because it allows the user or designer of data not to generate new datasets right away but instead create from existing ones with ease and much quicker than before.

Fault-Tolerant: It supports fault tolerance using RDDs which abstract the concept of worker nodes so that there will never be any lost work done with Spark; all actions are completed even if some part of a machine crashes or goes down unexpectedly.

Stream Processing: Spark supports stream processing in real-time. The earlier MapReduce could only process data already there and does not create new streams of information like Spark does.

Multiple-Language Support: With multiple languages like R, Scala and Python supported by Spark, it becomes easier for developers to write applications in their preferred language instead of having them stuck using only Java. This helps overcome limitations imposed due to Hadoop’s architecture that does not support any other programming languages apart from Java.

Hadoop Integration: Spark is a powerful tool that can be used to process large data sets. It integrates with other tools, including Hadoop, running on YARN clusters for increased scalability and reliability and support of Spark graphX for processing graphs efficiently in parallel without replicating information across nodes, which saves time.

Cost-efficient: The Apache Spark framework is considered to be a more cost-efficient solution when compared with Hadoop. Its ability to process data locally requires much less hardware than the traditional approaches of storing and retrieving large volumes from remote facilities or centralised clouds.

Developer’s Community: There have been many advancements in data analysis and management technology in the past few years. Apache Spark is considered one of the most important projects undertaken by this community because it offers a scalable system for processing large datasets with high-speed rates while also being easy enough that a developer can use it without much trouble.


  1. Explain RDD

The term “resilient distribution datasets” refers to RDDs. This is a fault-tolerant group of parallel operating components that can be split and divided among multiple nodes to improve scalability while staying immutable throughout the process.


There are two types: 

Paralellised collections, which means they’re meant to run simultaneously as one entity. These contain processing tasks like map reduce functions or merges between different sets–but sometimes even storage systems.

Hadoop Datasets (HDFS) are a type of big data that stores information on file systems like HDFS or other types. These perform operations for you in processing speeds to make your life easier.


  1. In Apache Spark, what is DAG?

Directed Acyclic Graphs (DAG) efficiently represent graphs that do not have directed cycles. There would be finite vertices and edges in a DAG, where each edge from one vertex points to another sequentially beside it. Operators represent work done on the data contained within these RDDs through its operations feature.


  1. What are the types of Deploy Modes in Spark?

Client Mode: When the spark driver component runs on the machine node from which the spark job is sent, it’s in client mode.

  • This approach has a major drawback – if the machine node fails, the whole operation fails.
  • The client mode enables you to use either interactive shells or job submission commands.
  • In a production environment, this client mode performs the worst and is not recommended.

Cluster Mode: In cluster mode, the deploy mode is said to be on a machine from which the spark job has been submitted if the spark job driver component does not operate on that machine.

  • The spark job starts the driver component within the cluster as a part of the ApplicationMaster sub-process.
  • The cluster mode supports deployment only with the spark-submit command (interactive shell mode is not accessible).
  • However, if the driver programs are run in ApplicationMaster fails, the driver program is re-created.
  • In cluster mode, a dedicated cluster manager (such as stand-alone, YARN, Apache Mesos, Kubernetes) is responsible for allocating the resources needed to execute the job. The following architecture depicts this.

Besides the client and cluster modes, there is the deployment mode used to deploy applications to our local machines for unit testing and development known as “Local Mode.” In this mode, the tasks are carried out on a single JVM in a single machine, making it highly inefficient since there will be a lack of resources at some point or another. It’s also not feasible to increase resources since the restricted memory and storage space limit that.


  1. In Apache Spark Streaming, what is your understanding of receivers?

Receivers in Apache Spark represent an abstraction for accessing cluster data. Any receiver can access all the data available on the cluster, which usually contains both streaming and batch data. 

Depending on how the data is sent to Spark, there are two types of receivers.

Reliable receivers send a copy of data to a given number of worker nodes. After receiving data from the data sources, the receiver sends an acknowledgement to the various data sources.

Unreliable receivers send the event data directly to the driver program and require developers to handle duplicates. There’s no notification provided to the data sources.


  1. What is the distinction between Coalesce and Repartition?


  • Spark Coalesce can only reduce the amount of data partitions.
  • Coalesce makes use of previously installed partitions to reduce the amount of randomly shuffled data.
  • Coalesce is faster than Repartition. However, if there are unevenly sized data partitions, performance may be somewhat reduced.



  • Repartition could decrease or increase the number of data partitions.
  • Repartitioning creates new data partitions and shuffles uniformly distributed data.
  • Repartition internally invokes Coalesce with the shuffle parameter, which makes it slower.


  1. List out the data formats supported by Spark.

Spark supports raw files and structured file formats for efficient reading and processing. File formats that Spark supports are Avro, CSV, JSON, RC, TSV, XML, paraquet, etc.


  1. What is Shuffling in Spark?

The term “shuffling” refers to the process of redistributing data across multiple partitions, either because it is required for JVM processes or executors on separate machines or simply to make use of the space more efficiently. The partition is a logical division of data that is smaller than the entire database. Do note that Spark has no control over which partition the data is split across.


  1. What is YARN in Spark?

Spark’s YARN is one of the important features that gives a centralised resource management infrastructure for distributing scalable operations across the cluster. It is a cluster management technology.

YARN distributes processing resources to different types of workflows by sending schedulers running on slave nodes in a Hadoop cluster, finding applications that are submitted to the ResourceManager for execution. Each application comprises one or more tasks, and each task is executed in its JVM. When an application needs access to the resources in a Hadoop cluster, it sends requests via YARN’s ApplicationMaster (AM) to the ResourceManager.



Apache Spark is a fast-growing cluster computational platform that was created to handle huge data more quickly while also supporting numerous big data applications and libraries. 

Connections like these provide the building blocks for both quick and robust applications, thanks to their use of many computational paradigms. Because of this, Spark has developed into a popular and lucrative technology, and understanding Spark will provide software developers and data engineers with access to new, better, and more difficult employment possibilities.

These are some of the frequently-asked Spark interview questions for new graduates and freshers that you may encounter while taking part in a Spark interview. 

Best wishes and good luck!