OK - which parts do you have issue with?
Have a look at this Momentum webinar ([Tech Brief On-Demand Webinar 2016] Provisioning with LANDESK Management Suite ) / grab the materials attached -- I explain in some detail how our PXE process works. Visual stuff may help I hope .
Having a clearer idea what you're struggling with would help.
Simple explanation:
- We only need a single PXE representative per broadcast domain (not the same as site). If you don't know what a broadcast domain is - essentially, anything that'll hear a UDP packet sent out on the local network is a broadcast domain ... that's USUALLY a site, but doesn't have to be. Note that this PXE Representative is (usually) a device with a LANDesk CLIENT installed - and NOT the Core Server!
- That PXE representative's job is to pick up the PXE - DHCP requests in each broadcast domain and fobs them off to the Core Server via civilised TCP and essentially goes "Hey - so I got this guy ... do I need to do anything with him?" ... and they handle it from there. So we're quite router friendly in that regard, since you don't NEED to do any UDP-forwarding or anything like that.
- The PXE reps are PROBABLY somewhat similar to WDS servers (in as far as - they pick up PXE requests) but they then pass them off to the Core Server (the central server), who then checks "do we have oustanding Provisioning jobs for those guys" and then passes a decision on to the PXE rep of the "Yes - tell this guy to boot into WIN PE and start template XYZ" or "Nope - nothing for this guy - he can boot into Windows". The PXE rep is pretty much "just" a middle man to make routers not be a "£$)(*" nightmare for the PXE process (which is not exactly operating on a robust protocol level) .
... it's really quite simple. PXE proxies make router configuration a non issue ... have one in each subnet that you're intending to use for OS Provisioning (that's where client self-electing services come in to make your life easier). And the PXE rops "simply" pass messages back & forth (in essence) between the client & the Core Server.
Does that make sense now?