A media company has been performing analytics on log data generated by its applications. There has been a recent increase in the number of concurrent analytics jobs running, and the overall…

Question

Options

• AUse a hash function to create a random string and add that to the beginning of the object prefixes
• BUse a lifecycle policy to change the S3 storage class to S3 Standard for the log data.
• CIncrease the read capacity units (RCUs) for the shared Amazon DynamoDB table.
• DRedeploy the EMR clusters that are running slowly to a different Availability Zone.

• A
  3% (1)
• B
  10% (3)
• C
  81% (25)
• D
  6% (2)

Explanation

When EMRFS consistent view is enabled, EMR uses a shared Amazon DynamoDB table to track S3 object metadata and ensure read-after-write consistency. As the number of concurrent EMR clusters and jobs increases, all clusters read from and write to this same DynamoDB table, which can exhaust the provisioned Read Capacity Units (RCUs). The symptom-task nodes taking longer to list S3 objects-is a DynamoDB throughput bottleneck, not an S3 prefix or storage class problem. Increasing RCUs gives the shared consistency table enough throughput to serve all concurrent clusters. Changing the S3 storage class (B) or moving to another AZ (D) would not resolve a DynamoDB capacity constraint, and a hash prefix (A) helps with S3 PUT throttling, not DynamoDB reads.

No community discussion yet for this question.