2015-11-03 20:07 GMT+01:00 Sebastian Kuepers <sebastian.kuep...@publicispixelpark.de>: > Hey, > > with collect() RDDs elements are send as a list back to the driver. > > If have a 4 node cluster (based on Mesos) in a datacenter and I have my > local dev machine. > > I work with a small 200MB dataset just for testing during development right > now. > > The collect() tasks are running for times faster on my local machine, than > on the cluster, although it actually uses 4x the number of cores etc. > > It's 7 seconds locally and 28 seconds on the cluster for the same collect() > job. > > What's the reason for that? Is that just network latency sending back the > data to the driver within the cluster? (well it's just this 200MB in total) > > Is that somehow a kind of 'management overhead' form Mesos? > > Appreciate any thoughts an possible impacts for that! Serialization and sending over network takes time, way more than simply processing the data on the same machine. But it doesn't scale as well. Try with more data and plot the results.
--------------------------------------------------------------------- To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
