Setting up a cluster with different memory sizes on nodes


I have cluster with 4 nodes having 20 GB RAM and 2 nodes having 40 GB RAM configured. However I don’t see that the 40 GB RAM nodes are getting more rows then the small nodes. Is this supposed to happen? Or is a heterogeneous setup not recommended because the small nodes are the limitations?

Thanks for your suggestions?


Aerospike evenly distributes data across the nodes, so the amount of data used on each node will be roughly the same. Also, work is proportional to the amount of data, which so is another reason to use homogeneous (in terms of memory, storage, chipset, networking).

If the only difference between those nodes is DRAM you can allocate the same memory-size for the namespace on all machines. If this is an in-memory namespace and your chipsets are different you should expect latency differences between those nodes as the load increases.


Thanks for clarification