Delevopment's list of Bittorrent

Justin Cormack | 1 Jun 2005 13:14

Re: using bittorrent for cluster file distribution


On 27 May 2005, at 02:17, larytet.8708132 <at> bloglines.com wrote:

> "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."
>
> do you have the source code
> opened ? i have tough problems with thread schedulers in the OSs. i  
> found
> out that its close to impossible to "sleep" for less than 10ms, for  
> example.
> it would be very interesting at least for me to discuss different  
> implementations
> of rate limiters.
>
> in my tests i reached 1500Kbyte/s (see for example http:// 
> larytet.sourceforge.net/images/tests/7clients_1300K.PNG)
> when running 7 clients on the same single CPU system. my current  
> implementation
> of the data trasnfer layer divides all available bandwidth by slots  
> of about
> 50Kbytes/s (or less). it means that to reach 15M between two peers  
> running
> on regular Windows host (all threads in Rodi client run in low  
> priority, besides
> socket Rx) i need ~30 up/down slots. In the terms of the Data  
> Transfer layer

(Continue reading)

headers

Tom H. Lan | 8 Jun 2005 07:50

question on Tit for Tat

Dear everyone,

I have got a qustion on Tit for Tat of BT. We know that peer will unchoke other peers if they can provide it with the best downloading rate.

Then does it mean that the uploading rate of a peer to its neighbor must be equal to the downloading rate it gets from its neighbor? It seems that if they are not equal, one may be choked.

Regards,

Tom

_______________________________________________
BitTorrent mailing list
BitTorrent <at> lists.ibiblio.org
http://lists.ibiblio.org/mailman/listinfo/bittorrent

headers

Olaf van der Spek | 8 Jun 2005 10:03

Re: question on Tit for Tat

On 6/8/05, Tom H. Lan <tomlan <at> ust.hk> wrote:
> Dear everyone,
>  
> I have got a qustion on Tit for Tat of BT. We know that peer will unchoke
> other peers if they can provide it with the best downloading rate. 
>  
> Then does it mean that the uploading rate of a peer to its neighbor must be
> equal to the downloading rate it gets from its neighbor? It seems that if
> they are not equal, one may be choked.

No, it doesn't.

headers

Olaf van der Spek | 8 Jun 2005 12:44

Re: question on Tit for Tat

On 6/8/05, Tom H. Lan <tomlan <at> ust.hk> wrote:
> So do you mean that the peer may offer more to its neighbor than it gets at
> some time? And then the peer may change to upload others by optimistic
> unchoking?

That depends on the choking algorithm. But the standard algorithm
indeed may, yes.

headers

Harold Feit - Depthstrike.com administrator | 8 Jun 2005 12:44

RE: question on Tit for Tat


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The likelyhood of a peer getting unchoked does increase if that peer
uploads more, and that peer is more likely to get upload as it
becomes available, but in top-heavy torrents (torrents with more
bandwidth supply than demand total) you'll see higher speeds on peers
that don't upload as much and peers that aren't reachable. Upload
rates don't need to be equal, but they should be proportional (under
ideal conditions).
 
The Tit for Tat enforcement is more visible on lower-bandwidth
torrents.

   _____  

From: bittorrent-bounces <at> lists.ibiblio.org
[mailto:bittorrent-bounces <at> lists.ibiblio.org] On Behalf Of Tom H. Lan
Sent: Wednesday, June 08, 2005 2:50 AM
To: bittorrent <at> lists.ibiblio.org
Subject: [bittorrent] question on Tit for Tat

Dear everyone,
 
I have got a qustion on Tit for Tat of BT. We know that peer will
unchoke other peers if they can provide it with the best downloading
rate. 
 
Then does it mean that the uploading rate of a peer to its neighbor
must be equal to the downloading rate it gets from its neighbor? It
seems that if they are not equal, one may be choked.
 
Regards,
 
Tom

- --
No virus found in this incoming message.
Checked by AVG Anti-Virus.
Version: 7.0.323 / Virus Database: 267.6.3 - Release Date: 6/6/2005

- -- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.323 / Virus Database: 267.6.3 - Release Date: 6/6/2005

-----BEGIN PGP SIGNATURE-----
Version: PGP 8.0.3

iQEVAwUBQqbMEF8nceBm0DUaAQJLLwf+M1xntb+fM56bpBlovnZn/uM4yczb+/gY
vxV57aCyZUwwxoN7hEU3vdQ5SSYU3iBMj4EEhRGJEZcOEv85AaQXAligJl39tTO+
rx02accL+kaxrQOvywrpwt/XnrPmwVdhBJ7yBQ9SWk2m+IBYPCCopHUvk85KmXha
uH3KddCwTJdyDcNgAVDpzBSCr71SVzcK9oXXi7p5+1u8LMIqdrJNgBLleQZajhCl
lOkVqDT3FX6JqPiAsrmtOjD/coHlFzJPmPtzEi/ox5vuv1+JWMXWG+Gc/RTyiK0Y
5fmmds8CkiH4bEt8GXbTpUbBJ39v1w7vnFLqdFCJwFXwT9W22o7i+g==
=uihU
-----END PGP SIGNATURE-----

headers

Tom H. Lan | 9 Jun 2005 14:38

a question on Random First Piece

Dear everyone,

I have a question puzzle me a lot and I need your help.

You see in BT that each file is chopped into different blocks and then further chopped into pieces. Peers relay the blocks between each other. I wonder whether block is the unit of the request. I mean that whether it is possible that peers will relay the block they are downloading which may have not been totally downloaded. If yes, it seems to me that the conception of block is useless.

And we know that a new peer who joins the BT networks will random choose a block to download. Does it mean that it can only upload to others when it finishes downloading the block.

Thanks very much.

Bests,

Tom

_______________________________________________
BitTorrent mailing list
BitTorrent <at> lists.ibiblio.org
http://lists.ibiblio.org/mailman/listinfo/bittorrent

headers

Justin Cormack | 9 Jun 2005 15:00

Re: a question on Random First Piece


On 9 Jun 2005, at 13:38, Tom H. Lan wrote:

> Dear everyone,
>
> I have a question puzzle me a lot and I need your help.
>
> You see in BT that each file is chopped into different blocks and  
> then further chopped into pieces. Peers relay the blocks between  
> each other. I wonder whether block is the unit of the request. I  
> mean that whether it is possible that peers will relay the block  
> they are downloading which may have not been totally downloaded. If  
> yes, it seems to me that the conception of block is useless.
>
> And we know that a new peer who joins the BT networks will random  
> choose a block to download. Does it mean that it can only upload to  
> others when it finishes downloading the block.

Yes thats right, peers will only request a block when the other peer  
sends a 'have' message, which will only be when the whole block has  
been downloaded and verified, so there will not be requests for  
unverified parts.

Justin

headers

Simon Spero | 9 Jun 2005 16:57

Re: a question on Random First Piece

The sub-piece/block mechanism is *fatally* flawed; you should never take *any* action based on a info in a partial block.

Unfortunately the flow-control algorithm means you have to use partial blocks for bandwidth calculations, but you shouldn't send out things to your peers until you've validated it.

Simon

On Jun 9, 2005, at 8:38 AM, Tom H. Lan wrote:

Dear everyone,

I have a question puzzle me a lot and I need your help.

You see in BT that each file is chopped into different blocks and then further chopped into pieces. Peers relay the blocks between each other. I wonder whether block is the unit of the request. I mean that whether it is possible that peers will relay the block they are downloading which may have not been totally downloaded. If yes, it seems to me that the conception of block is useless.

And we know that a new peer who joins the BT networks will random choose a block to download. Does it mean that it can only upload to others when it finishes downloading the block.

Thanks very much.

Bests,

Tom

_______________________________________________

BitTorrent mailing list

BitTorrent <at> lists.ibiblio.org

http://lists.ibiblio.org/mailman/listinfo/bittorrent

_______________________________________________
BitTorrent mailing list
BitTorrent <at> lists.ibiblio.org
http://lists.ibiblio.org/mailman/listinfo/bittorrent

headers

larytet.8708132 | 10 Jun 2005 23:17

MD5 collision attack

it's getting hotter
http://www.cits.rub.de/MD5Collisions/

esp idea of
colliding X.509 certificates

headers

Tom H. Lan | 11 Jun 2005 05:21

question on the blocks downloading in BT

Dear everyone,

You see when a peer is downloading a block, the uploader of the bolck may depart the networks anytime. So it is very likely that a peer have some partially downloaded blocks. How peers deal with these blocks? Do they ask other peers to upload these block based on what they have? Or just drop these blocks?

Many Thanks.

_______________________________________________
BitTorrent mailing list
BitTorrent <at> lists.ibiblio.org
http://lists.ibiblio.org/mailman/listinfo/bittorrent

2	January 2013
1	December 2012
2	August 2012
3	May 2012
1	April 2012
2	December 2011
5	November 2011
1	October 2011
2	September 2011
14	August 2011
10	July 2011
8	June 2011
2	May 2011
3	April 2011
4	February 2011
3	January 2011
7	October 2010
6	July 2010
13	June 2010
8	April 2010
2	March 2010
22	January 2010
2	December 2009
14	November 2009
6	September 2009
5	August 2009
16	June 2009
13	May 2009
13	April 2009
5	March 2009
25	February 2009
1	January 2009
15	December 2008
5	August 2008
8	July 2008
7	June 2008
4	March 2008
19	February 2008
3	January 2008
2	December 2007
9	November 2007
12	October 2007
17	September 2007
40	August 2007
3	July 2007
4	June 2007
7	May 2007
34	April 2007
17	March 2007
7	January 2007
22	December 2006
31	November 2006
24	October 2006
19	September 2006
19	August 2006
20	July 2006
46	June 2006
25	May 2006
34	April 2006
57	March 2006
23	February 2006
14	January 2006
26	November 2005
47	October 2005
56	September 2005
11	August 2005
37	July 2005
63	June 2005
60	May 2005

Mon	Tue	Wed	Thu	Fri	Sat	Sun

		1	2	3	4	5
6	7	8	9	10	11	12
13	14	15	16	17	18	19
20	21	22	23	24	25	26
27	28	29	30