NBMA and Broadcasts

Neeko

I've been going over frame relay and OSPF recently and am unclear on NBMA networks and broadcasts.

First off, when we talk about broadcasts, what actually determines the broadcast addressing? I'm assuming it's different depending on what layer 2 link is used. For example on an Ethernet link, I'd assume broadcast addressing would be required at layer 2 and possibly 3. For example ARP, isn't this unicast at layer 3 but broadcast at layer 2? Does ARP even use an IP header for that matter?

What about over a serial link using PPP or HDLC? Would this just use layer 3 broadcast addressing?

Also, does layer 3 addressing change from a segment based broadcast such as a DHCP request (I wont use ARP as an example since I'm not sure of the layer 3 addressing here), to an all segments based broadcast such as RIP? I'd guess the subnet broadcast address would be used for a segment e.g 192.168.10.255 /24 and type of broadcast used by RIP for example, would be 255.255.255.255 therefore being sent out of all interfaces with IP enabled. Is this right?

Moving on... so with frame relay, why is it actually referred to as non broadcast and multi access? The multi access part I'd say refers to the ability to reach multiple destinations via a single interface or subinterface.

With what I was saying about differences between broadcast addressing depending on the layer 2 link, the non broadcast part I'd assume refers to the inability to send a layer 3 broadcast out of a single interface or subinterface with multiple VCs configured on it, and have that packet broadcasted at layer 2 in order to reach the peers on the other end of each VC.

Please tell me I'm close lol.

OSPF, network types and Chris Bryant (CCIE number 12933 as if I could ever forget ) has opened up a can of worms and in typical CCNA fashion, raised many questions for the eager learner without answering them. The ICND2 exam cert guide doesn't even mention it, but it's too late I can't leave it now. Before getting straight to the OSPF over frame relay questions I thought it would be best if I get all this underlying broadcast stuff clear conceptually.

kryolla

Let me try and explain it

You ARP when you dont know the layer 2 MAC address on an ethernet so the packet looks like this
Source IP address is known and Source MAC address is known, Destination IP address is known and Destination MAC is all 1s. The host with that IP address will reply wth its own MAC address as the source. So for an ARP packet the only broadcast is layer 2. RIPv1 on ethernet sends a broadcast at layer 2 and 3 on UDP port 520 so every router listening on port 520 will pick up the RIP packet. NBMA for frame relay means you can have more than 2 nodes on the same subnet whether it is a full mesh or not. Non-broadcast means there is no broadcast DLCI or layer 2 address. So for a full mesh when you ping the broadcast it will go out all DLCI that are active for a hub and spoke it will stop at the hub.

Hope this helps

Neeko

Oops I edited the original post slightly before I realised anyone had replied.

I understand what ARP acheives, but I just checked wireshark and couldn't see IP on the ARP entries, it jumps straight from the Ethnernet frame to Address Resolution Protocol. Looks to me like there is no IP header, is this because the ARP info defines the IP dest and source and is only meant for the local link? DHCP would include an IP header though and that could be meant for the local link, or is that different because it comes from the application layer and need encapsulation at each lower layer?

I understand when you refer to no layer 2 broadcast for frame relay, but not sure what you mean by stopping at the hub?

Thanks for the reply.