AWS Official Blog

New Instance Types for Amazon Elastic MapReduce

by Jeff Barr | on | in Amazon Elastic MapReduce | | Comments

Thousands of AWS customers use Amazon Elastic MapReduce to process and store vast amounts of data. Because Elastic MapReduce is built around the Hadoop framework, it is easy to use hundreds or thousands of Amazon EC2 instances in parallel.

Hot on the heels of the price reductions that we made last week, we are also adding support for 12 additional EC2 instance types:

  • General Purpose – m3.xlarge and m3.2xlarge.
  • Compute-optimized – c3.xlarge, c3.2xlarge, c3.4xlarge, and c3.8xlarge.
  • GPU – g2.2xlarge.
  • Memory-optimized – cr1.8xlarge.
  • Storage-optimized – i2.xlarge, i2.2xlarge, i2.4xlarge, and i2.8xlarge.

We are also adding support for existing instance types in additional AWS Regions:

  • hs1.8xlarge is now available in five additional Regions (US West (Oregon), Europe (Ireland), Asia Pacific (Tokyo), Asia Pacific (Singapore), and Asia Pacific (Sydney)).
  • hi1.4xlarge is now available in two additional Regions (US West (Oregon) and Asia Pacific (Tokyo)).
  • cc2.8xlarge is now available in one additional Region (Asia Pacific (Tokyo)).

Here are some specs and hints that will help you to choose a suitable instance type for your own jobs:

  • The hs1.8xlarge instance type is ideal for large long running clusters that need lots of local disk, making it ideal for a data warehouse based on Hadoop.  It comes with 16 vCPUs, 117 GiB of memory, 24 x 2 TB local storage, and 10 Gigabit network connectvity.
  • The c3.8xlarge instance type is ideal for compute-intensive workloads, perfect for image processing and other CPU-bound workloads.  It comes with 32 vCPUs, 60 GiB of memory, 2 x 320 GB SSD storage, and 10 Gigabit network connectvity.
  • The i2.8xlarge instance type is packed with CPU, memory, and storage, and is ideal for analytics applications like Impala, Spark, and HBase.  It comes with 32 vCPUs, 244 GiB of memory, 8 x 800 GB SSD, and 10 Gigabit network connectvity.
  • The m3.2xlarge instance type provides a good balance of compute, memory, and network resources. This instance is perfect for many standard workloads, including those that do not need a lot of local storage. It has 8 vCPUs, 30 GiB of memory, 2 x 80 GB SSD storage, and fast networking. 

— Jeff;