Hi, Do the two nodes with the highest load reside on the same physical machine? If that is so and the queries were run from that machine, then I could assume that the cpu load is due to the reduce step being held there.
Kind regards, Alex 3 апр. 2017 г. 8:45 PM пользователь "vdpyatkov" <vldpyat...@gmail.com> написал: Hi, How was Ignite nodes and physical machines mapped? Where Ignite nodes more there is CPU utilization greater, isn't it? Where were SQL query generated (on one of a srever node or other)? I think, CPU utilization shuold be same over of nodes, because SQL executed as Map Reduce task. -- View this message in context: http://apache-ignite-users. 70518.x6.nabble.com/CPU-Usage-on-distributed-queries-tp11652p11680.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.