I have a question regarding how the PCIe topology is created for virtual machines. By topology I mean how the root complex connects to endpoints via PCIe switches: the number of switches, which endpoint connects to which port of the switch and the entire tree that is created by endpoints and switches. So the question is: Is the PCIe topology always guaranteed to be the same for a given hardware configuration?
I see that in practice that if I have a hardware configuration (say 2 SCSI controllers, 3 VmNet3 NIC cards, etc) the PCI addresses (bus:device.function) is the same everytime I boot and also everytime I create a VM with the same harware configuration. Is this guaranteed to be the case always? Can the PCI addresses change if I create another VM?
I have an application wherein the NIC cards are identified by the PCI address (bus:device.function). If the PCI addresses change from boot to boot or even from VM to VM, the application will not work as intended.
Another question regards OVFs. If VMs are deployed using OVFs, will the PCIe topology be the same everytime? Will it be exactly the same as the VM the OVF was created from?
I have searched the forums but could not get the answer to this question. Any help would be greatly appreciated.