3
  1. I came up with this idea of using more than LAN cable for transmitting data. Suppose we have 6 wires running between two clients, we can divide the each packet into 4 equal parts (2 for error detection and correction), and transmit it simultaneously. On receiving the packet, the receiver combines all the fragments to get back the packet.
  2. Also, if 6 cables seem to much, we can use 2 cables, both for only one way transmission. This way, the receiver doesn't have to send acknowledgement via the same channel, and can help de-stress the connection.

How are the ideas?

7
  • 1
    That's a lot of wiring, don't you think? You'd greatly increase the cost of hardware (you need 6 interfaces instead of 1) as well as the cost of running all that extra cable everywhere. – Ron Trunk Oct 26 '15 at 13:42
  • 1
    Also, in full-duplex switched Ethernet, there are separate channels for sending and receiving. – Ron Trunk Oct 26 '15 at 13:43
  • In certain scenarios where speed is more important than cost, can it practically implemented? I am trying work on a paper on this. Thanks for the help. – Atul Goel Oct 26 '15 at 13:47
  • 2
    It already is implemented. Google "link aggregation" and take a look at IEEE 802.3ad. – Ron Trunk Oct 26 '15 at 13:49
  • 2
    802.3ad does not fragment a frame. Each frame will flow through one, and only one, link. The hash function used to select a link will usually keep a connection flowing across the same link, 'tho there are systems that will "round-robin" frames across all available links. – Ricky Oct 26 '15 at 21:24
2

This is not done in the general case because of error handling and reassembly issues due to differing transmission delays. Plus, the full frame would have to be received to divide it up for parallel transmission -- store-and-forward switching is generally avoided due to the latency it introduces.

Some switches (eg. Cabletron) can do this via proprietary mechanisms. In the cabletron case, it requires the parallel links be equidistant to within a few feet; if a fragment is damaged in transit, the reassembled frame will contain that damage. (i.e. they didn't bother with error detection/correction)

0

In a sense we have done this. If you look at something like 100GBASE-SR10 you find that at the physical layer it is essentially a group of ten 10GBASE-SR links.

It's generally better to do this stuff at as low a level as possible because any buffering or delay mismatches between the communication streams will make the reassembly process much harder.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.