Why a short ping and a long UDP Moscow - Helsinki?

Question

Computer networks

Short TCP / IP and ping between Moscow and Helsinki take an average of 40ms. A long UDP block of 12288 bytes has a delay of up to a second. Does anyone know the cause and how to fix it? If you cut the block, then to what length?

Answer 1

The reason is the size of the packet, which must first be fragmented along the way, and glued back together in place. The dumber the piece of iron, the more difficult it is for it. If it is not possible to configure this at the application / OS level, you can try to throw a tunnel that will do this by itself, sending packets of a normal size to the Internet. Standard - 1500 (1460) bytes.

Answer 2

There are two reasons:
1. The UDP protocol does not have a packet acknowledgment mechanism, it was conceived as an alternative to TCP for those cases when acknowledgment is not needed. If the packet is lost along the way, it is simply retransmitted (if upper layer protocols detect its absence, of course). It is because of this that a characteristic "gurgling" occurs during negotiations on SIP phones. If the packet is lost, then the upper-level protocol can request it again (if it keeps track of the sequence itself) and until it receives it, it will be hammered
2. Fragmentation on devices during transmission. It is enough to get caught on the whole route alonedevice with a small MTU - and that's it, sailed, the packet will be broken into parts that will fit into it. The piece of iron for transmission allocates a buffer no less than a packet in size, the piece of iron for transmission too and starts accepting pieces until it receives the entire packet (if something went wrong, everything starts all over again). The larger the packet, the greater the risk that it will be damaged along the way and retransmitted.