Hadoop vs Spark comparison - Practice Questions

Challenge - 5 Problems

🧠 Conceptual

intermediate

Key difference in data processing models

Which option correctly describes the main difference between Hadoop MapReduce and Apache Spark in terms of data processing?

A Both Hadoop MapReduce and Spark process data only in real-time streaming mode.

B Spark uses disk storage for all intermediate data, whereas Hadoop MapReduce keeps data in memory throughout the process.

C Hadoop MapReduce processes data in batches using disk storage between steps, while Spark processes data in-memory for faster computation.

D Hadoop MapReduce is designed for in-memory processing, while Spark is optimized for batch processing on disk.

💡 Hint

Think about how each system handles intermediate data during processing.

Hadoop MapReduce writes intermediate data to disk between map and reduce steps, which slows down processing. Spark keeps data in memory, making it faster especially for iterative tasks.

🧠 Conceptual

intermediate

Fault tolerance mechanisms

Which statement best explains how fault tolerance is handled differently in Hadoop and Spark?

A Hadoop uses data replication on HDFS for fault tolerance, while Spark uses lineage information to recompute lost data.

B Spark replicates data blocks across nodes like Hadoop to ensure fault tolerance.

C Both Hadoop and Spark rely solely on checkpointing to handle faults.

D Hadoop does not support fault tolerance, but Spark does through data replication.

💡 Hint

Consider how each system recovers lost data after a failure.

Hadoop relies on replicating data blocks on HDFS to recover from failures. Spark tracks transformations (lineage) and can recompute lost data without replicating it.

❓ data_output

advanced

Comparing execution time for iterative tasks

Given the following scenario: A machine learning algorithm runs 10 iterations on a large dataset. Which framework will likely complete the task faster and why?

A Both will have similar speeds because they use the same storage system.

B Spark, because it caches data in memory across iterations, reducing disk I/O.

C Hadoop MapReduce, because it writes intermediate results to disk for each iteration.

D Hadoop MapReduce, because it uses in-memory caching by default.

💡 Hint

Think about how iterative algorithms benefit from in-memory data storage.

Spark caches data in memory, so iterative algorithms run faster by avoiding repeated disk reads/writes. Hadoop MapReduce writes intermediate data to disk each iteration, slowing down the process.

🧠 Conceptual

advanced

Resource management differences

Which option correctly describes how Hadoop and Spark manage cluster resources?

A Spark manages resources internally without any external cluster manager.

B Spark only runs on Hadoop YARN and cannot use other resource managers.

C Hadoop and Spark both use the same resource manager exclusively called Mesos.

D Hadoop uses YARN for resource management, while Spark can run on YARN, Mesos, or standalone cluster managers.

💡 Hint

Consider the flexibility of Spark in cluster environments.

Hadoop uses YARN as its default resource manager. Spark is more flexible and can run on YARN, Mesos, or its own standalone cluster manager.

🚀 Application

expert

Choosing the right framework for a real-time analytics project

You need to build a system that processes streaming data from sensors in near real-time and performs complex analytics. Which framework is best suited and why?

A Spark, because it supports real-time streaming with low latency and in-memory processing.

B Hadoop MapReduce, because it is optimized for real-time streaming data processing.

C Neither Spark nor Hadoop can handle streaming data; a traditional database is better.

D Hadoop MapReduce, because it processes data faster by using disk-based batch jobs.

💡 Hint

Think about which framework supports streaming and fast analytics.

Spark supports real-time streaming with Spark Streaming and Structured Streaming APIs, enabling low-latency analytics. Hadoop MapReduce is batch-oriented and not designed for real-time streaming.