Mark Lu created GIRAPH-1016:
-------------------------------
Summary: Number of Workers and Giraph Speed
Key: GIRAPH-1016
URL: https://issues.apache.org/jira/browse/GIRAPH-1016
Project: Giraph
Issue Type: Task
Affects Versions: 1.1.0
Environment: aws ec2 Linux.
Reporter: Mark Lu
I am trying to run giraph's SimpleShortestPathsComputation to processing a
small graph dataset with nearly 77510 vertices and 898900 edges on aws ec2
instances, (T2.micro with 1 master and 2 slave nodes), Hadoop version is 1.2.1.
The giraph command is
hadoop jar giraph-with-dependencies.jar org.apache.giraph.GiraphRunner
org.apache.giraph.examples.SimpleShortestPathsComputation -vif
org.apache.giraph.io.formats.JsonLongDoubleFloatDoubleVertexInputFormat -vip
/user/ec2-user/a2.txt -vof
org.apache.giraph.io.formats.IdWithValueTextOutputFormat -op
/user/ec2-user/output1 -w 1.
As I increase the number of workers (ie, -w 2,3...), the cpu time as well as
the total time of giraph computation is also increased. So should the cpu time
and computation time decreased when more workers are added? What should I do?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
