Bitzu

SAA-C03 ยท Elastic compute performance

AWS Batch versus Amazon EMR

Running large batch jobs with AWS Batch versus big data processing with Spark or Hadoop on EMR, and when Fargate or Lambda is enough.

Key points

  • AWS Batch: containerized batch jobs, no scheduler to manage.
  • Job (job definition: image, vCPU, memory) goes to a job queue; Batch scales a compute environment: EC2 On-Demand or Spot, Fargate, ECS Managed Instances.
  • Amazon EMR: Spark, Hadoop, Hive; primary node coordinates, core nodes run tasks and store HDFS, task nodes only compute (Spot).
  • Transient clusters auto-terminate, data in S3.
  • Lambda: short (under 15 min) event-driven work.

Source: https://docs.aws.amazon.com/batch/latest/userguide/batch_components.html

Practice questions

Single choice

In AWS Batch, where do you submit a job so that it waits until it can be scheduled onto compute resources?

  1. A job definition
  2. A compute environment
  3. A job queue
  4. A scheduling policy
Show answer

Answer: C. A job queue

Jobs are submitted to a job queue, where they wait until the scheduler places them on a compute environment linked to that queue. The job definition is the blueprint, the compute environment is the capacity, and a scheduling policy only controls fair-share allocation within a queue.

Select all that apply

Which statements about Amazon EMR node types are correct?

  1. The primary node coordinates work and tracks cluster health
  2. Core nodes run tasks and store data in HDFS
  3. Task nodes store HDFS replicas for durability
  4. Task nodes are required in every cluster
  5. A cluster can run with no primary node
Show answer

Answer: A. The primary node coordinates work and tracks cluster health, B. Core nodes run tasks and store data in HDFS

The primary node coordinates the cluster and tracks task status, and core nodes both run tasks and store data in HDFS. Task nodes are optional and do not store HDFS data, so they can be removed or interrupted without losing HDFS blocks. Every cluster has a primary node, and a single-node cluster is possible.

True or false

AWS Batch can only run jobs on EC2 instances, so serverless container capacity such as AWS Fargate is not an option for a Batch compute environment.

Show answer

Answer: False

False. A managed AWS Batch compute environment can use Fargate, Amazon EC2 (On-Demand or Spot) or Amazon ECS Managed Instances. Fargate removes instance management when jobs fit its resource limits.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More SAA-C03 topics in this module