I'm wondering if your 'Deploying in a Cluster' is right

Because when I run a small dataset using als and sgd with 8 slaves, I noticed that each node load the whole file. Whether it worked or not is really bothering me. Could you help me?