MapReduce quiz

The map function can output 0/1/more values for each key-value pair

• True

• False

The architecture of Map Reduce is decentralized to operate in parallel

• True

• False

If the memory allocated to a certain map task runs out, the whole thing fails

• True

• False

What is the worst case complexity of the shuffling phase (in the number of tasks) ?

• O(1)

• O(n)

• O(n log n)

• O (n^2)

A regular MapReduce (v1) job can take multiple hours to complete. Why?

• Jobs usually involve large datasets and complex computations that require substantial time to distribute, process, and aggregate data across multiple nodes.

• Some machines might become "stragglers" and we have to wait for them.

• There might be multiple rounds of shuffling data, which can create network bottlenecks and delay the processing time.

• The execution of tasks is sequential and thus we lose a lot by not parallelizing.

If a failure happens during the reduce phase (v1), it is necessary to start the job over entirely

• True

• False

How many files a single reduce task can produce by default?

• 0

• 1

• many

What are the necessary properties of a reduce function, that make the reuse of it as a combine function possible?

• Commutativity

• Distributivity

• Associativity

• Matching of the intermediate and output formats

Which of the following can exist?

• Map phase

• Combine slot

• Reduce function

• Map slot

• Combine function

The input key/value pairs of map functions are sorted by the key

• True

• False