Hi Mimmo, So I think the fastest way is to send compute jobs to all nodes which store data and in those compute jobs run scan queries for local partitions and save files to local file. Also you can scan different partition in parallel to utilize all available CPUs. In this case you won't have any contention for data access and utilize all CPUs in the grid.
if you need to gather data on one particular node, you can do the same but without compute job. Just run scan queries for by partitions in parallel to utilize all CPUs. Thanks, Mikhail. -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/