Question about L2/L3 and routing

[Jan Dvorak]

Hello,

theoretically, is the cisco switch (6500 family in particular) able to route packets within one L2 segment (one vlan) using default gateway instead of direct MAC?

So we have computer A with MAC 00:00:00:00:00:AA, computer B with MAC 00:00:00:00:00:BB and the default gateway for this vlan with MAC of 00:00:00:00:00:DD.

Packet has to be sent from A to B, but here for whatever reason computer A will be using DGW MAC (00:00:00:00:00:DD) instead of directly B's computer MAC (00:00:00:00:00:BB).

I know that this behavior makes no sense, but in theory (in case of wrong node A behavior), is there any standard (rfc?) which does explicitely prohibits this behavior? I belive the switch/router should still be capable of delivering the packet from A to B, right?

thanks for help,

BR, Jan

[rikshit4aggarwal]

You may use a hub or a small switch if you directly want to go from A to B..If you want to use bigger switches, you may have to configure the ports in the same VLAN. You just need to configure the IP addresses in the same subnet ,the switch will learn about both the computers using ARP and it should work even without a default gateway. 

Regards,

Rikshit

[Jan Dvorak]

Hi Rikshit,

thanks. Yes, I do understood that, arp for mac and then send directly to mac of second node. But the question is just a generic - in same vlan, two nodes - and for some reason the node A is NOT using ARP of node B, but instead is sending the packets to DGW from the same vlan. Should this work or is it probihited (by any rfc/standard)?

As I have seen a node behaving like that, sending to DGW and the switch/router was sending packets to node B, but once the traffic increased, the packets start to be dropped (eg 20% of them).

Need a conclusion if this a failure of the switch/router or if it's now allowed to route like that in L2 network segment and this situation had never occur.

thanks,

Jan

[Joseph W. Doherty]

I know that this behavior makes no sense, but in theory (in case of wrong node A behavior), is there any standard (rfc?) which does explicitely prohibits this behavior?

Not that I'm aware of.

I belive the switch/router should still be capable of delivering the packet from A to B, right?

A packet or a frame?

If we're only dealing with a frame, the receiving host would consider the frame delivered and wouldn't know what else to do with it.

If we're dealing with a packet, assuming it correctly has host B's IP, a router would route the packet to host B.  As a router would be routing back onto the same network (assuming hosts A and B are in the same network - being within the same VLAN doesn't guarantee that), and assuming redirects aren't disabled, the router would send a redirect message to host A (basically telling host A, it can send traffic directly to host B without using the default gateway - which normally it shouldn't if it "knows" it and host B are in the same network).

[Jan Dvorak]

thanks Joseph.

A packet or a frame?

It were packets, just was using DGW..

assuming hosts A and B are in the same network

Yes, same VLAN and also same subnet

outer would send a redirect message

I did not saw any redirect packets while analyzing the issue, but can't confirm on that (as I do not admin the router).

the behavior of the router was that it was handling the packets fine even using DGW and not directly host B, but once the traffic increased there was a packet loss introduced (actually quite significant one)..

thanks & regards,

Jan

[Joseph W. Doherty]

Well on routers, before CEF, there was an interface command for caching when the ingress and egress interface were the same.  You're running CEF, but if you're not:

IP Input

The Cisco IOS^® software process called IP input takes care of process-switching IP packets. If the IP input process uses unusually high CPU resources, the router is process-switching a lot of IP traffic. Check these issues:

• If an interface has a lot of secondary addresses or subinterfaces and there is a lot of traffic sourced from the interface and destined for an address on that same interface, then all of those packets are process-switched. In this situation, you should enable ip route-cache same-interface on the interface. When Cisco Express Forwarding switching is used, you do not need to enable Cisco Express Forwarding switching on the same interface separately.