I'm using Apache toree kernels( Scala, PysparK ) to submit a spark job on yarn remotely.
-Edge Node contains ( Jupyter Notebook ) on Azure 4cores, 14Go ram.
-MapRSandBox 5.2 on Cloud Azure: 4 cores, 14Go ram.
it's take too long to create a sqlcontext from edge node than directly on MapRSandBox !!
Is it a question of performance ??