Big Data Integration and Processing
Quiz 5 Answer
Quiz 5 - Pipeline and Tools
Q1) What is data-parallelism as defined in lecture?
- Having multiple multiple data pipelines at the same time.
- Simultaneously processing input data from multiple cores.
- Running the same function simultaneously for the partitions of a data set on multiple cores.
- At each step of the data pipeline, process values simultaneously by using multiple cores.
Q2) Of the following, which procedure best generalizes big data procedures such as (but not limited to) the map reduce process?
- split->sort->merge
- split->do->merge
- split->map->shuffle and sort->reduce
- split ->shuffle and sort->map->reduce
Q3) What are the three layers for the Hadoop Ecosystem? (Choose 3)
- Data Creation and Storage
- Data Management and Storage
- Data Integration and Processing
- Data Manipulation and Integration
- Coordination and Workflow Management
Q4) What are the 5 key points in order to categorize big data systems?
- Coordination, Latency, Productivity, Speed, Fault Tolerance
- Execution model, Speed, Scalability, Flexibility, Fault Tolerance
- Coordination, Latency, Productivity, Flexibility, Fault Tolerance
- Execution model, Latency, Scalability, Programming Language, Fault Tolerance
Q5) What is the lambda architecture as shown in lecture?
- A type of swappable data processing layer.
- An architecture that natively supports lambda calculus.
- A type of hybrid data processing architecture.
- A type of architecture that only contains part of the data processing method.
Q6) Which of the following scenarios is NOT an aggregation operation?
- Removing undefined values.
- Counting the total number of data.
- Counting the total number of data per type.
- Averaging the total number of data per type.
Q7) What usually happens to data when aggregated as mentioned in lecture?
- Data become organized.
- Data becomes smaller.
- Data becomes personalized.
- Data becomes faster to process.
Q8) What is K-means clustering?
- Classify data by k actions.
- Divide samples using k lines.
- Classify data by k decisions.
- Group samples into k clusters.
Q9) Why is Hadoop not a good platform for machine learning as mentioned in lecture? (Choose 4)
- Too massive.
- Java support only.
- Bottleneck using HDFS.
- Map and Reduce Based Computation.
- Unable to support machine learning.
- No interactive shell and streaming.
- Requires nodes and multiple machines.
Q10) What are the layers (parts) of Spark? (Choose 5)
- MLlib
- Graphx
- SparkSQL
- Spark Graph
- Spark Core
- Spark RDD
- Spark Streaming
- Worker Node
Q11) What is in-memory processing?
- Having the input completely in disk.
- Having the input completely in memory.
- Having the pipeline completely in disk.
- Having the pipeline completely in memory.
- Writing data to disk between pipeline steps.
- Writing data to memory between pipeline steps.
--------------------------------------------------------------------------------------------------------------------------------------------------------
Big Data Integration and Processing
0 Comments