fragmentation

Question

Why has fragmentation  done on routers only . As I know the reason of doing fragmentation  is data link layer can only receive 1500 bytes  (standard Ethernet) from Upper layer  (network layer ) so if the upper  layer data is bigger  than 1500 bytes than it will be fragmented  so that data link layer can hold the data for further process it as a frame .

So a/c to me fragmentation  should be done at the source  end itself a/c to its MTU  . Why  is it done on router? ?

Please help me out . Thank you

Answer 1

This is because the data gram has to travel to different networks and those different network may have different Maximum Transmissible Unit (MTU). The source (sender) can only know about the MTU of the current network and Network Layer at the source can fragment accordingly. But when the packet travels in the inter-networks, the MTU might be different. Since routers are the one responsible for connecting networks, so they are the best ones to do the fragmentation, when needed.

Comments

@arunav

 yes but still I have to clear one thing that   in that case I have mentioned  (2000 byte of ip Packet and MTU is 1500 only ) if i am sending this packet  to some other network  whose MTU is 100 byes supposed   then will it get fragmented   2 times one at the source and other one at the router.  

( I know these r silly  doubts, but I am very  much  confused. Clear it please )

---

@bharti

Yes it would be fragmented whenever the MTU of the other network where datagram has to travel is lesser.

So, in your case, the fragmentation will happen twice. However, in case of TCP, the TCP layer uses Path MTU discovery and fragments the packet accordingly. The TCP layer at sender would find out, using Path MTU discovery, that on the way to the destination, the lowest MTU is 100 bytes (in your case). So it would send send those size packets (100 - 20 bytes header) to the IP layer.

More at: https://en.wikipedia.org/wiki/Path_MTU_Discovery

---

Thank you  @ arunav  it helped  me really.   :)