[cid:[email protected]]<http://www.amazon.com/gp/r.html?R=2Y2HNBKSM1BG9&C=1W26C1FKPTSLH&H=OWAPTU86BT440OWXHPCAIVQ1NSUA&T=C&U=http://aws.amazon.com?ref_=pe_2170_16291170>
We are excited to announce the immediate ability for Amazon Elastic 
MapReduce<http://aws.amazon.com/elasticmapreduce/> customers to dynamically 
modify the size of their running job flows. Through the API, command line 
client, or Java SDK, you now have the flexibility to add or remove slave nodes 
based on the changing capacity needs of your job flow. You may also include a 
predefined step in your workflow that automatically resizes a job flow between 
steps that are known to have different capacity needs. This will allow you to 
better customize your capacity usage to meet the unique requirements of your 
job flows, improving performance while enabling you to save money.
Two types of slave nodes are now supported: core nodes and task nodes. Core 
nodes can be added to a running job flow but cannot be removed and should be 
reserved for the capacity that is required until your job flow completes. Task 
nodes, which can be both added and removed from a running job flow, are ideal 
for the capacity that is only needed on a temporary basis. These nodes can be 
of different EC2 instance types, giving you additional flexibility into the 
composition of your cluster. Additionally, core nodes both host persistent data 
using Hadoop Distributed File System (HDFS) and run Hadoop tasks, while Task 
nodes only run Hadoop tasks.
Elastic MapReduce also now automatically provisions new slave nodes if a slave 
node fails, providing you an even greater level of fault tolerance.
Please refer to the Resizing Running Job 
Flows<http://docs.amazonwebservices.com/ElasticMapReduce/latest/DeveloperGuide/ResizeJobFlow.html>
 section in the Elastic MapReduce Developer's Guide for more details. Note that 
you will have to download the newest version of the Ruby 
CLI<http://developer.amazonwebservices.com/connect/entry.jspa?externalID=2264> 
to use this feature.

Sincerely,

The Amazon Elastic MapReduce Team

Reply via email to