DAS-C01 · Question #73
A financial company uses Apache Hive on Amazon EMR for ad-hoc queries. Users are complaining of sluggish performance. A data analyst notes the following: - Approximately 90% of queries are submitted…
The correct answer is D. Create instance group configurations for core and task nodes. The two key observations point to a scaling problem: (1) 90% of queries arrive at the same time (bursty workload), and (2) HDFS utilization stays below 10%, indicating data lives in Amazon S3 (EMRFS), not HDFS-meaning compute, not storage, is the bottleneck. Instance group…
Question
A financial company uses Apache Hive on Amazon EMR for ad-hoc queries. Users are complaining of sluggish performance. A data analyst notes the following:
- Approximately 90% of queries are submitted 1 hour after the market
opens.
- Hadoop Distributed File System (HDFS) utilization never exceeds 10%.
Which solution would help address the performance issues?
Options
- ACreate instance fleet configurations for core and task nodes.
- BCreate instance fleet configurations for core and task nodes.
- CCreate instance group configurations for core and task nodes.
- DCreate instance group configurations for core and task nodes.
How the community answered
(43 responses)- A14% (6)
- B9% (4)
- C5% (2)
- D72% (31)
Explanation
The two key observations point to a scaling problem: (1) 90% of queries arrive at the same time (bursty workload), and (2) HDFS utilization stays below 10%, indicating data lives in Amazon S3 (EMRFS), not HDFS-meaning compute, not storage, is the bottleneck. Instance group configurations support Auto Scaling policies on task nodes, allowing EMR to automatically add task nodes when the query surge hits and remove them afterward. Instance fleets (choices A/B) also support scaling but use a different allocation strategy (mix of instance types/Spot). The exam distinguishes these: instance groups with Auto Scaling are the recommended pattern for predictable, time-based spikes. Creating task node groups specifically allows horizontal scale-out without touching core nodes, which manage HDFS data. Since HDFS is barely used, scaling task nodes (pure compute) is the right lever.
Topics
Community Discussion
No community discussion yet for this question.