[bittorrent] using bittorrent for cluster file distribution

Justin Cormack justin at street-vision.com
Thu May 26 17:48:42 EDT 2005


On Thu, 2005-05-26 at 20:06, Ryan Stenson wrote:
> Hi all,
>  
> I'm a consultant at the Cornell Theory Center (www.tc.cornell.edu
> <http://www.tc.cornell.edu/> ).  We're considering using BitTorrent to move
> large files from a file server/job submit node to the compute nodes (1:M).
> Using multicast, we get about 30-40 Mb/s, however our network (GigE) should
> be capable of 800+.  Does anyone have similar experience with this sort of
> application?
>  
> I've setup Bittorrent on a few servers in an attempt to speed up our current
> multicast file distribution system (we do a lot of copying from a few
> fileservers to dozens of clients).  
>  
> I start the tracker with:              bttrack.py --port 6969 --dfile dstate
> I then create the torrent with:     btmaketorrent.py http://IP:6969/announce
> c:\temp\filename
> and seed the torrent with:          btdownloadheadless.py
> \\server\filename.torrent --save_as c:\temp\filename
> and download it to the clients with the same command.
>  
> I'm only seeing 20 KB/s transfer speeds over Gigabit ethernet (these
> machines are dual xeon's, 4gb ram, etc.).  We see 30-40 Mb/s over multicast,
> which is extremely slow compared to our theoretical throughput of 800+ Mb/s.
> Are there any default settings that are limiting my throughput?  Any other
> suggestions?  I realize that using the start bittorrent code, uncompiled,
> isn't the optimal solution, but this is just the first step. 

Well a few things.

This has come up a few times recently, but havent really had much
feedback from anyone. Have you got some time to run some tests?

First, I am writing a BT client for this type of application. The first
version (possibly a few residual bugs, although seems pretty reliable)
does work. I didnt do any real performance tests, although on the same
host with a large file, moving it to the same hard drive I got around 20
MB/s, so one should be able to do much better between machines.

I am rewriting the new version in my (non-copious) spare time, and its a
bit more portable (first version is Linux only), although I dont have
any current intentions of porting it to Windows which you seem to be
using.... (unless you pay me...)

However you could do some helpful tests:

First check transfer on the same machine (preferable different disks)
Then check between 2 machines only, across the network
Check your network performance (I use ttcp for Unix, should be able to
compile it for windows)

Then once you have done that theres more, depending if there really is a
point to point bottleneck, or whether it is a multipoint bottleneck.

Oh, and what switches are you using and how many machines are there?

j





More information about the BitTorrent mailing list