cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
566
Views
0
Helpful
10
Replies

MTU question

waimen
Level 1
Level 1

Hi All,

In the following case:

--R1--R2--

MTU of R1 is 1500 and MTU of R2 is 1000. So if R1 send a packet of 1500 bytes to R2, will R2 drop the packet? What should I do in this scenario? Thank You!

Waimen

10 Replies 10

rjackson
Level 5
Level 5

Technically MTU is a "transmit" limit, it should not affect receiving larger packets.

t.baranski
Level 4
Level 4

I don't know the theoretical answer to the question "what does a station do if it receives a packet larger than its MTU?", but... The MTU is the maximum size segment that can be transmitted on a physical medium. With this in mind, how can R1 and R2 -- if they are indeed directly-connected as per above -- have different MTUs unless something has been misconfigured?

How about the MSS value? I think they are related with each others.

Regards

ahvn
Level 1
Level 1

Hi,

I too feel what rjackson in his reply told was right. MTU denotes the Maximum transmission unit in that case there should not be any problem for the incoming traffic to the router. And I think instead of keeping the value 1000 for MTU you can change that to 1500.

Rgds,

Homin

michacarter
Level 1
Level 1

mtu refers to the maximum that can be transmitted before fragmentation occurs.

Be aware of other issues, such as if you run ospf - you will not get a neighbour if the mtu is mismatched. Also, with ip multicast enabled, interfaces which transmit traffic from larger to smaller mtu interfaces may not be hardware switched due to fragmentation - degradation in performance may occur (note: this refers to interfaces in the same device).

If an interface receives a packet which is greater than its configured MTU, it will drop the packet and report that as an input error - giant packet.

So if you have mismatched MTU's, you might run into issues later on. Also let me ask you this, is there any specific reason why you need to do so? just wondering....

Anup

thought that when an interface recieves a packet larger than its MTU size, it will fragment it, then send it. But if the "do not fragment bit" has been set on the packet, it will be dropped. In this case, it will then send an ICMP message back to the sender, for it to reduce its MTU size

This is the case if the packet is larger than the MTU of the interface that the router is going to forward the packet out of (the outbound interface). But it's not clear what happens if the packet is larger than the MTU of the interface on which the packet is received.

I believe there are two cases to consider: 1) the case where the MTU of an interface is artificially lowered below the maximum for some reason (e.g., the use of IPSEC/GRE tunnels), and 2) the case where the MTU is set at the maximum for the interface's technology (e.g., 1500 for Ethernet).

In case 1, we know that the interface can accept the packet from a layer-1 standpoint as long as the packet doesn't exceed the maximum MTU of the network technology in use. So it seems logical to say that in this case the interface would be more likely to accept such a packet. I verified that my workstation (WinXP) behaves this way -- I've set the MTU at 1400 because I'm behind a tunnel, and the interface accepts (and responds to) 1500 byte ICMP pings.

Case 2 is another story. Here you're sending a packet that's larger than what the receiving interface was designed to accept. My guess is that the results would be implementation-dependant, and that the odds of success would be inversely porportional to the difference between the packet's size and the interface's MTU. (When VLAN tags were invented, Ethernet's maximum packet size increased by 4 bytes (to 1522 bytes). It turned out that virtually all existing interfaces were able to accept packets of the new maximum size even though they were only designed to accept packets up to 1518 bytes in size.)

Then again, when would case 2 ever occur in a properly-behaving network? (MTU-increasing features from IEEE aside ;-)

We have experienced problems when using both VLAN tagging and tag-switching (MPLS) on an ethernet interface.

How should we configure the mtu to make this combination work ?

l.bookstaver
Level 1
Level 1

The MTU on both ends should always be the same. If there is a need for one end to have a smaller MTU, the "ip mtu 1000" would be appropriate rather than just the "mtu 1000" on the primary interface. Here is my example:

interface Serial0/0

no ip address

encapsulation frame-relay IETF

no ip mroute-cache

!

interface Serial0/0.100 point-to-point

ip address 172.16.1.105 255.255.255.252

ip mtu 800

frame-relay interface-dlci 100

With mismatched physical MTUs, I would expect you would find some things work fine and some don't. My experience was for example ftps worked but Exchange clients would hang.

My real world example is what I expect is a malfunctioning CSU or bad cable. In bound packets to the router of any size work fine, but if I use the default mtu of 1500 for outbound packets, a lot of the larger packets will just disappear. Using an "ip mtu 800" on my interface works around the issue.

The local router will drop the large packets being sent out the serial interface and send the icmp back to the the original local host indicating it needs the DF bit clear. So everything worked even though the clients would have negoiated a larger TCP MSS.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: