Using your own partitioner didn't work?

e.g.
YourRDD.partitionBy(new HashPartitioner(your number))

xj @ Tokyo

On Wed, Sep 10, 2014 at 12:03 PM, qihong <qchen@pivotal.io> wrote:
I'm working on a DStream application.  The input are sensors' measurements,
the data format is <sensor id><timestamp><measure>

There are 10 thousands sensors, and updateStateByKey is used to maintain
the states of sensors, the code looks like following:

val inputDStream = ...
val keyedDStream = inputDStream.map(...)  // use sensorId as key
val stateDStream = keyedDStream.updateStateByKey[...](udpateFunction)

Here's the question:
In a cluster with 10 worker nodes, is it possible to partition the input
dstream, so that node 1 handles sendor 0-999, node 2 handles 1000-1999,
and so on?

Also, is it possible to keep state stream for sensor 0 - 999 on node 1, 1000
to 1999 on node 2, and etc. Right now, I see sensor state stream is shuffled
for every batch, which used lot of network bandwidth and it's unnecessary.

Any suggestions?

Thanks!



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/how-to-setup-steady-state-stream-partitions-tp13850.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
For additional commands, e-mail: user-help@spark.apache.org