Hi, Here are my responses:
1. In standalone mode each executor will run an embedded client node to access the cluster. If you use saveValues() method to load the data, IgniteRDD will generate keys in the way that values will be saved on the closest server (i.e. running on the same box with the executor). This should minimize data movement. 2. For this you will have to use Ignite API directly. You can use IgniteContext.ignite() to get the local client on the driver. -Val -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Apache-Spark-Ignite-Integration-tp8556p9006.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.
