如何使用 UDP 协议构建可靠的 P2P 数据传输

Question

as you can read in the title I have to build a reliable P2P data transfer Using UDP(I'm a student), so what I asking you guys is to not give me the code, actually, I hate copying and pasting so much, I feel pain in doing it, I'm asking for help to tell me what tool do I need, I'm familiar with Javascript and Java and Python, feel free to help me if you know any of these languages.

the reliability part will be achieved through checksum and ACKs(Acknowledgements), so I have to Implement them: -I know how checksum can be calculated. -ACKs can be Implemented in the way of Stop-and-wait protocol(I think it's the simplest one), if you know another protocol, that is okay.

I'm really lost, I don't know from where I begin if you have some code examples, please share a link that would be helpful, so I can build an idea from where do I start.

thanks.

Answer 1

Here are some reliable UDP examples in github like kcp, quic, utp. You can read them and get the essential from them.

As for how to do it, there are some little suggestions:

1. You need ACK to check if any packet/datagram lost, and then retransmit it
2. You need to use FEC
3. Congestion algorithm can be considered later.

When you try to build your own code, do a little goal in every step, don't build a complex protocol at the very begining.

Answer 2

If you want an explanation - here is the resource

From one student to another - my.networking class used Kurose and Ross, and their Chapter 3, section 3.4 on reliable data transfer breaks this down in a really good manner.

They start by assuming Reliable Data Transfer through the.network, with an 'rdt 1.0' protocol to transfer through that.network, and slowly add in the different types of unreliability (such as packet loss or packet corruption). As you reach 'rdt 3.0', you should have what you want - reliable data transfer on top of UDP. Ofcourse, their approach is a bit more abstract, but it is a good explanation. I can link in their final sender and receiver state machine, and the name of the textbook.

Here is a breakdown of what the textbook explains:

1. Start by assuming a completely reliable.network. You just need to send the packets from the sender to the receiver.

2. Assume the.network now has bit corruption. You will need to transmit a checksum to allow for recovery incase the packet was corrupted. This also requires ACK (and initially NAK but this is eliminated with sequence numbers on the ACK). Remember, your ACK are also packets going through the.network, and they can also get corrupted! (the solution was using sequence numbers with the ACK)

3. Assume the.network can now lose packets. The reliable data transfer now needs a timeout mechanism to retransmit a packet if it's ACK was not received.

If you want just the implementation - Here is the resource

Textbook - Kurose and Ross, Computer Networking, A Top-Down Approach Chapter - 3, Section - 3.4

The receiver state machine was left as an exercise to the reader, but should be easy to pick up from their explanation.

Note - I am in no way affiliated with the textbook mentioned in this question. I just used it for a.networking class and I felt it had a good explanation for the question asked.