We are planning to upgrade our small office virtualization infrastructure to hyperconverged cluster. Currently failover is manual, using Hyper-V replication feature.
I would like a second opinion, if we have understood everything correctly.
There are 3 nodes running free Hyper-V 2016 Core (hardware is HP ML310e gen8, HP ML10 gen9 and Fujitsu RX1330 M4)
We are planning to install Intel X520-DA2 NICs(two SPF+ ports) or, better yet, Intel XL710 quad port (4 SPF+ ports) in all hosts. And connect them using SPF+ direct attach cables and fiber (one of the nodes is in another rack room for safety)
Trouble is, we only have 3 fiber ports available on each end, so we can connect directly either StarWind sync or iscsi/heartbeat, but not both. Which of those are more demanding? The other one would need to be switched through the remaining port.
Also both sync and iscsi/heartbeat would run on the same hardware. Or would it be better to get two, separate 2 port NICs instead of one 4 port? Not sure if all of the servers support two PCI-E 3.0 8x and such approach is more expensive. We could provide additional heartbeat channel via some VLAN through the regular network NIC.
I have adapted a schema to illustrate the situation. Red line represents the gap between rack rooms, which are connected with said 3 fiber lines. Example has iSCSI as direct connections, and sync via switch, but it could be vice versa.
Also green part is redundant or does it provide any benefit?
Will all of this work with free Hyper-V Server license?
Will this work with StarWind VSAN Free (which is limited to 3 nodes as I understand)?
Maybe it is cheaper/easier to buy some of-the-shelf storage solution for our setup?
The Latest Gartner® Magic Quadrant™Hyperconverged Infrastructure Software