Research faster spin up times

[sbernauer]

When creating a HDFS cluster the data-nodes will get created one after each other.
Compared to hbase where all region-servers are started in parallel this takes quite some time if the number of datanodes is large.
We should research if there is the need to start all datanodes sequentially or if we can improve the startup time somehow