I use Spark2 on Yarn. I using ORC Files with partitions and caching in memory. I testing concurrent request of Spark.
If just only user request, Spark very fast, take 2-3s to finish.
But when I increase in few user request, Spark seem very slow, It take 10-15s to finish request.
So I need to be able to handle many more requests.
Anyone have any document or practical about this.
