A question often fielded by ConfigMgr OSD implementers is how an organization can
have multiple PXE providers on the same network subnet. Usually, the client organization
already has a PXE server in place to support a legacy imaging product. The real answer to
this question is that it depends on the network infrastructure, not on ConfigMgr.
PXE is a standards-based protocol based on DHCP and network broadcasts. The network
card installed on a system controls the actual booting of a system; this is completely
independent of ConfigMgr. Generally, the first PXE provider to respond to a PXE request is
chosen by the network card that is booted. On a network level, it is best to segregate PXE
providers on separate subnets to control the broadcasts.
You can also specify a specific PXE server using DHCP options 60, 66, and 67; however,
these options are specific to a single network subnet and cannot be made more granular
for PXE booting purposes. There are other options including a newer PXE specification, but
this problem is completely outside the bounds of ConfigMgr and WDS.
Two excellent resources for detailed PXE information are at http://technet.microsoft.com/en-us/library/cc732351.aspx and http://support.microsoft.com/kb/244036 . The second reference is a dated KB article referring to RIS, but the general PXE information is still valid.