Would someone have an idea about how much of system overhead (cpu, memory including NVRAM) MCC-IP will take compared to a standalone ?
Of course it can vastly be dependent upon the ISL network, but without that put into perspective, meaning it is all optimal, only thinking system overhead taking place just due to the MCC operation, how much more should I assume ?
in general Metrocluster functionality is something that is just part of ONTAP itself. So mainly write IO has to taken into account when talking about overhead since that has to be written twice incl. commitment from the remote site due to the Sync Mirror underlying functionality. I guess you´re talking asking about a new install, setup or prospect on-prem engineered system, right? Hence backend covered by 10/40 or 100G Ethernet/IP. Since you asking about NVRAM mirroring - that adds the only (one known to me) aspect that has to taken into account since it has to be mirrored both ways too BUT most implementation will use dedicated HW components to take care especially of that load. I suggest to get in contact with your local or our (based in Germany) Solutions Architects team with long term MC experiences.