I believe that the correct answer to your question - don't do that. The strength of distributed systems is that you have a number of identical pieces which you can scale out virtually with no limits.
If your cluster is heterogenous - i.e. all the nodes are different in size, amount of data and power - you may easily run into issues with uneven load distribution and eventually into performance issues. Even more, you can run into cluster management issues, e.g. if one big node leaves the cluster then the remaining small nodes will redistribute the data and may not have enough space to do that. If you really-really want something like that, and you think that your specific use case is not going to suffer from uneven load or cluster management issues, then go with the suggestion from Stephen - run multiple nodes on big machines. Stan On Fri, Nov 1, 2019 at 4:36 AM BorisBelozerov <arsenal18...@gmail.com> wrote: > Can you give me some code to implement it? Thank you!! > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >