file swarming?????????????????
 

I am just wondering...

How does p2p networks implement multi-sourced download, i.e. downloading the same file from multiple sources.

P2P is something that realy interest me- I am doing a small research about it

Now, I understand UDP would be a very good protocol for implementing this. but still as a programmer I have many questions.

Anyone who knows a good site that explains how does multi-sourced download works - please give me a link.

its the only thing missing for me to completly understand how p2p works.

Many thanks in advance,
SIJP

Hi sijp and welcome to P2P-Zone.

The typical way to handle multisourced downloads is to treat each shared file as a collection of regular size chunks. For example, you could have a 5 MB file split into 50 chunks, each 100 kB in size. When you have downloaded all the chunks belonging to a particular file, you can join them together and voilà - you have a complete file. So instead of requesting a whole file from a sharer you request only the chunks you need. And you can download different chunks of the same file simultaneously from different people, which usually results in a faster total download speed.

As you may receive chunks for a large file from hundreds of different sources, data integrity becomes an important issue. This is typically handled with a method called hashing. A specific, hard-to-forge hash number is calculated for each chunk belonging to a file, and the downloaders can use this number to check that they have received a chunk with correct content. If some source seems to send a lot of false chunks, the p2p program can block that source as a hostile peer and get the chunks from other sources.

- tg :WA:

You can easily see what TankGirl said by going into an Overnet temp folder. You'll see a movie broken down into 75 parts for example.

What I want to know is how WinMX handles its multisource transfers. Even when recieving from ten differents sources the file always downloads from begining to end, not in random chunks; it acts like a single source download. It does seem to use a bit more upstream bandwidth during downloads and lots of processor time too; mayby it makes dozens of 5K or 10K transfer requests per second and keeps a close watch to make sure the chunks download in the right order.