Run kmeans on Twister

1. Prepare 1.1 to start activemq and twister in advance.

xiuwenyi@Twister1:~/apache-activemq-5.4.2/bin$ ./activemq console

xiuwenyi@Twister1:~/twister-0.9/bin$ ./start_twister.sh 

1.2 jar package (on all nodes) Run ant under $ twister_home/samples/kmeans to copy the generated Twister-Kmeans-0.9.jar to the $ twister_home/apps path

2. Run the kmeans algorithm 2.1 to create the kmeans input directory

xiuwenyi@Twister1:~$ cd $TWISTER_HOME/bin xiuwenyi@Twister1:~/twister-0.9/bin$ ./twister.sh mkdir kmeans

2.2 generate data

xiuwenyi@Twister1:~/twister-0.9/bin$ cd $TWISTER_HOME/samples/kmeans/bin xiuwenyi@Twister1:~/twister-0.9/samples/kmeans/bin$ ./gen_data.sh init_clusters.txt 3 3 /kmeans km_data 6  60000

Automatically upload the generated data files to each node in the cluster on average. If you do not need to generate data automatically, use the following command to allocate data.

cd $TWISTER_HOME/bin ./twister.sh put $TWISTER_HOME/samples/kmeans/input kmeans

2.3 create a partition File

xiuwenyi@Twister1:~/twister-0.9/samples/kmeans/bin$ cd $TWISTER_HOME/bin xiuwenyi@Twister1:~/twister-0.9/bin$ ./create_partition_file.sh kmeans km_data $TWISTER_HOME/samples/kmeans/bin/kmeans.pf

2.4 run

xiuwenyi@Twister1:~/twister-0.9/samples/kmeans/bin$ cd $TWISTER_HOME/samples/kmeans/binxiuwenyi@Twister1:~/twister-0.9/samples/kmeans/bin$ ./run_kmeans.sh init_clusters.txt 6 kmeans.pf