Hi everyone.
I'm a grade 9 learner with a science expo due for 15 June 2011 (2 weeks from now)
I seriously need to start and was thinking about doing a project where I perform an investigation to decide whether websites should use either the Bittorrent protocol or HTTP protocol to host their downloads.
My hypothesis is: Bittorrent should be used for the larger files (200 mb+) and HTTP should be used for basically any files under 200 mb.
I could perform the investigation by analyzing how much traffic went to the host server when downloading through Bittorrent and when downloading through HTTP - this should basically prove that hosting files with Bittorrent is cheaper since it is a lot less harsh on the server receiving the download requests.
The other part of my investigation will be a speed comparison between Bittorrent and HTTP where I will just see how fast different sized files are downloaded when being downloaded from either the Bittorrent protocol or HTTP protocol.
The problems I face:
1. Should I add more protocols to my comparison?
2. How will I upload the torrent? (ie: through a torrent site or using a tracker program and bittorrent)
3. How will I get a good amount of seeders on the torrent? (Maybe I should find a commercial website which offers free http and torrent downloads for all their files then just use that.)
4. How will I be able to see how much traffic goes to the hosting network of the file(s)?
5. What other tests/investigations should I perform? (I can't do the whole expo on just speed and traffic...)
Thanks!
Comparing Bittorrent to HTTP
Moderators: kgudger, bfinio, Moderators
Re: Comparing Bittorrent to HTTP
Hi fuz,
There is no single good answer to your question. As you suspect, the problem with bittorrent is that with few or no seeders, performance will be really bad, especially since the hosting sites tend to have a pretty poor bandwidth in practice. HTTP is good, and FTP is probably better, and you can even use the "ftp://" designator to make the web browser use it. You will still have the problems of 1) a hosting site with good bandwidth, and 2) potential degradation of performance if many people try to download at the same time.
It also depends on what you want to do. Distributing a zipped data file is different from streaming a video because there are specialized software packages for video, some of which are cleverer than HTTP or FTP at dealing with bandwidth issues. This is an area of active research, and there are several commercial and open source projects that take a variety of strategies.
Doing this for free is different from doing it as a business. If you are doing it for money and have a budget, you can use the services of, for example, Akamai. That company has a sophisticated worldwide network of about 100,000 servers dedicated to distributing digital content in the most efficient way practical. Many companies, e.g. Apple's iTunes, use them for heavy data distribution.
Getting back to your hypothesis, you would need to test it by varying the number of seeders, such as by coordinating with your friends to have some of them seed at specific times while others do several test downloads (at least three or more attempts per file to get valid statistics). Having the load distributed that way will probably give you more credible data than if you do it all yourself at home or on computers that are all on the same local network. You would probably also want to test different torrent sites to host the files, or just stick to one that is reliable. You could vary the torrent client, but that would add too many variables to your experiment, I think. If you compare with FTP and HTTP, preferably on the same server, you will probably get some pretty good numbers.
If you are knowledgeable enough, you might be able to set up your own server with bittorrent, HTTP, and FTP, and have your friends point to it. That way you can directly compare the performance of each service while eliminating the variable of whether they are running on different servers or not. You could try setting up a linux box (e.g. with Ubuntu) and setting up your home router to allow ftp, http, and bittorrent traffic to it (and only to it! this is risky!). This is somewhat sophisticated stuff, and you need to be careful that you don't let the evil hacker hordes into your home network. Remove the hole as soon as you are done with your tests. If you can do this on a computer that is already visible on the internet and not in your home, so much the better. You should assume that bad people will quickly try to get into your network and do destructive and illegal things.
Let me know what you think and how you think you can go about it.
There is no single good answer to your question. As you suspect, the problem with bittorrent is that with few or no seeders, performance will be really bad, especially since the hosting sites tend to have a pretty poor bandwidth in practice. HTTP is good, and FTP is probably better, and you can even use the "ftp://" designator to make the web browser use it. You will still have the problems of 1) a hosting site with good bandwidth, and 2) potential degradation of performance if many people try to download at the same time.
It also depends on what you want to do. Distributing a zipped data file is different from streaming a video because there are specialized software packages for video, some of which are cleverer than HTTP or FTP at dealing with bandwidth issues. This is an area of active research, and there are several commercial and open source projects that take a variety of strategies.
Doing this for free is different from doing it as a business. If you are doing it for money and have a budget, you can use the services of, for example, Akamai. That company has a sophisticated worldwide network of about 100,000 servers dedicated to distributing digital content in the most efficient way practical. Many companies, e.g. Apple's iTunes, use them for heavy data distribution.
Getting back to your hypothesis, you would need to test it by varying the number of seeders, such as by coordinating with your friends to have some of them seed at specific times while others do several test downloads (at least three or more attempts per file to get valid statistics). Having the load distributed that way will probably give you more credible data than if you do it all yourself at home or on computers that are all on the same local network. You would probably also want to test different torrent sites to host the files, or just stick to one that is reliable. You could vary the torrent client, but that would add too many variables to your experiment, I think. If you compare with FTP and HTTP, preferably on the same server, you will probably get some pretty good numbers.
If you are knowledgeable enough, you might be able to set up your own server with bittorrent, HTTP, and FTP, and have your friends point to it. That way you can directly compare the performance of each service while eliminating the variable of whether they are running on different servers or not. You could try setting up a linux box (e.g. with Ubuntu) and setting up your home router to allow ftp, http, and bittorrent traffic to it (and only to it! this is risky!). This is somewhat sophisticated stuff, and you need to be careful that you don't let the evil hacker hordes into your home network. Remove the hole as soon as you are done with your tests. If you can do this on a computer that is already visible on the internet and not in your home, so much the better. You should assume that bad people will quickly try to get into your network and do destructive and illegal things.
Let me know what you think and how you think you can go about it.
Heinz Hemken
Mentor
Science Buddies Expert Forum
Mentor
Science Buddies Expert Forum