I personally feel that this should be handled on higher level. In my view, Ignite Compute subsystem is about simple delivery of a job to a remote server. It has failover and all, but to really use that you have to go for ComputeTaskSplitAdapter or similar things, and they’re hard to get right.
I’d rather see some sort of distributed job scheduler as a higher level subsystem, e.g. built as an Ignite Service. Or, one could say that all you need is a distributed priority queue. We already have a regular FIFO queue as an Ignite Data Structure. Why not add another one? In any case, I think you should ask on [email protected] what other people think. And the best way to get this feature is to contribute it! :) Stan From: chewie Sent: 16 октября 2018 г. 11:31 To: [email protected] Subject: RE: Priority on affinityCalls Hi Stan, Thanks for your quick response! Could something like prioritized callables be considered in a future version? Currently in my project a bunch of client nodes send all kinds of affinityCalls to the server nodes and a real-time interaction task and a slow moving batch job probably shouldn't be executed with the same urgency. I just feel I wouldn't be the only one gaining from such a feature. /Anders -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
