headers
Detlef Bosau | 1 Jan 2007 13:48
Picon

Re: Are we doing sliding window in the Internet?

Fred Baker wrote:
>
>
> I wonder where you got the notion that a typical session had a 10 ms 
> RTT. In a LAN environment where the servers are in the same
It´s just a number.  However, it´s the magnitude. Not the exact number. 
In your example below you have about 
20 ms,
100 ms,
200 ms
RTT.

If we again consider one outstanding segment with 12000 bit we have the 
following rates (approximately)
600 kbps
120 kbps
60 kbps

It´s not the question whether this is optimal. It´s the question: Does 
this happen in a relevant number of cases?
Particularly in downloads from mobile devices which quite often do not 
offer larger bandwidth.

So to your question why TCP should tune itself to only one outstanding 
segment: The reason could be the limited bandwidth the node, e.g. a 
mobile node, can handle.

Detlef
(Continue reading)

Permalink | Reply | 0 comments |
headers
Detlef Bosau | 1 Jan 2007 13:49
Picon

Re: Are we doing sliding window in the Internet?

Medhi, Deep wrote:
> See 
>
> John Heidemann, Katia Obraczka, and Joe Touch. "Modeling the Performance of HTTP Over Several Transport
Protocols." ACM/IEEE Transactions on Networking, vol. 5, pp. 616-630, October, 1997. 
>
> This covers maximum usable window size for different transmission media.
>
> 	-- Deep
>   
Unfortunately, I don´t have an ACM account. Is it possible to send me a 
copy? Perhaps Joe?

Thanks a lot!
Permalink | Reply | 1 comment |
headers
Detlef Bosau | 1 Jan 2007 17:36
Picon

Thanks a lot for the copies! Re: Are we doing sliding window in the Internet?

I just received two copies of the paper. Many thanks to all!

Detlef

Detlef Bosau wrote:
> Medhi, Deep wrote:
>> See
>> John Heidemann, Katia Obraczka, and Joe Touch. "Modeling the 
>> Performance of HTTP Over Several Transport Protocols." ACM/IEEE 
>> Transactions on Networking, vol. 5, pp. 616-630, October, 1997.
>> This covers maximum usable window size for different transmission media.
>>
>>     -- Deep
>>   
> Unfortunately, I don´t have an ACM account. Is it possible to send me 
> a copy? Perhaps Joe?
>
> Thanks a lot!
>
>
Permalink | Reply | 0 comments |
headers
Fred Baker | 1 Jan 2007 01:29
Picon
Favicon

Re: Are we doing sliding window in the Internet?

yes and no.

A large percentage of sessions are very short - count the bytes in  
this email and consider how many TCP segments are required to carry  
it, for example, or look through your web cache to see the sizes of  
objects it stores. We are doing the sliding window algorithm, but it  
cuts very short when the TCP session abruptly closes.

For longer exchanges - p2p and many others - yes, we indeed do  
sliding window.

I don't see any reason to believe that TCPs tune themselves to have  
exactly RTT/MSS segments outstanding. That would be the optimal  
number to have ourstanding, but generally they will have the smallest  
of { the offered window, the sender's maximum window, and the used  
window at which they start dropping traffic }. If they never see  
loss, they can keep an incredibly large amount of data outstanding  
regardless of the values of RTT and MSS.

I wonder where you got the notion that a typical session had a 10 ms  
RTT. In a LAN environment where the servers are in the same building,  
that is probably the case. But consider these rather more typical  
examples: across my VPN to a machine at work, across the US to MIT,  
and across the Atlantic to you:

[stealth-10-32-244-218:~] fred% traceroute irp-view7
traceroute to irp-view7.cisco.com (171.70.65.144), 64 hops max, 40  
byte packets
1  fred-vpn (10.32.244.217)  1.486 ms  1.047 ms  1.034 ms
2  n003-000-000-000.static.ge.com (3.7.12.1)  22.360 ms  20.962 ms
(Continue reading)

Permalink | Reply | 155 comments |
headers
Venkata Pingali | 2 Jan 2007 19:31
Picon
Favicon

Re: Are we doing sliding window in the Internet?


A few months back we collected some
per-connection data in both client
and server modes. We thought you might
be interested in the preliminary results.

We collected data in two modes/configurations.
In the client mode we configured Apache to
be a web proxy  and in the server mode we
configured Apache to serve an actual website.
The basic results, which must be only considered
as being indicative/hints of the reality, are
as follows:

Server end (i.e, end that has large
amount of data to transfer):

     - Most connections are short (90% < 1sec)
     - MaxCwnd is < 5KB in > 80% of cases
     - MaxRTT is distributed almost uniformly
       in the 0-400ms range.

Client end (i.e., the end receiving data):

     - ~ 90% of connections see MaxCwnd < 5KB
     - < 1% connections see MaxCwnd > 10KB
     - 90% of connections have MaxRTT < 100ms

There are some problems with the data:
(Continue reading)

Permalink | Reply | 153 comments |
headers
Detlef Bosau | 2 Jan 2007 20:52
Picon

Re: Are we doing sliding window in the Internet?

Venkata Pingali wrote:
>
>
> Server end (i.e, end that has large
> amount of data to transfer):
>
>     - Most connections are short (90% < 1sec)

Do you have any knowledge of the number of "rounds" the TCP connection 
has seen?  A couple of years ago I saw some similar result (don´t no the 
source at the moment) where 90 % of connections consist of not more than 
20 packets.

Now, consider the initial slowstart, IIRC we start with 2 MSS (?) then 
we have:

Round    CWND
    1              2
    2              4
    3              8
                            total of 14 packets up to now
    4             16
                            total of 24 packets up to now,

thus many flows will finisch before the end of the fourth round which 
would correspond to a CWND of about 6 kByte, 1500 byte MSS assumed.

In short words: Quite a few connections are finished before the end of 
the fist slow start period.
(Continue reading)

Permalink | Reply | 152 comments |
headers
Venkata Pingali | 2 Jan 2007 21:29
Picon
Favicon

Re: Are we doing sliding window in the Internet?

Detlef Bosau wrote:
> Venkata Pingali wrote:
>>
>>
>> Server end (i.e, end that has large
>> amount of data to transfer):
>>
>>     - Most connections are short (90% < 1sec)
> 
> Do you have any knowledge of the number of "rounds" the TCP connection 
> has seen?  A couple of years ago I saw some similar result (don´t no the 
> source at the moment) where 90 % of connections consist of not more than 
> 20 packets.

Our sample shows that 94% of connections
have < 20 packets - when observed from the
server end.

     Number of Packets       Percentile of Connections

      3                         4%
      4                         55%
      5                         69%
      10                        87%
      20                        94%

I have included the new graph and generated pdfs.

http://www.isi.edu/aln/e2e.pdf
http://www.isi.edu/aln/e2e.ppt
(Continue reading)

Permalink | Reply | 0 comments |
headers
Joe Touch | 3 Jan 2007 01:14
Picon
Favicon

Re: Are we doing sliding window in the Internet?


Detlef Bosau wrote:
> Venkata Pingali wrote:
>>
>>
>> Server end (i.e, end that has large
>> amount of data to transfer):
>>
>>     - Most connections are short (90% < 1sec)
> 
> Do you have any knowledge of the number of "rounds" the TCP connection
> has seen?  A couple of years ago I saw some similar result (don´t no the
> source at the moment) where 90 % of connections consist of not more than
> 20 packets.
> 
> Now, consider the initial slowstart, IIRC we start with 2 MSS (?) then
> we have:

I don't know if the current code starts with 2 MSS; it could start with 4.

> Round    CWND
>    1              2
>    2              4
>    3              8
>                            total of 14 packets up to now
>    4             16
>                            total of 24 packets up to now,

It doesn't double each RTT; it goes up by 50%. Remember, the window
grows by one MSS each ACK during the initial phase, but there is one ACK
(Continue reading)

Permalink | Reply | 150 comments |
headers
Joe Touch | 3 Jan 2007 03:55
Picon
Favicon

Re: Are we doing sliding window in the Internet?


Lachlan Andrew wrote:
> Greetings,
> 
> On 02/01/07, Joe Touch <touch <at> isi.edu> wrote:
>>
>> Detlef Bosau wrote:
>> > Round    CWND
>> >    1              2
>> >    2              4
>> >    3              8
>>
>> It doesn't double each RTT; it goes up by 50%. Remember, the window
>> grows by one MSS each ACK during the initial phase, but there is one ACK
>> for each two MSS's.
> 
> If you have ABC (as recent Linux senders do by default), or don't use

ABC is EXPERIMENTAL.

> delayed ACKs (as Linux receivers don't when the window is small),

Delayed ACKs are strongly encouraged.

Both good reasons to fix these bugs in Linux.

> Detlef was right that it doubles each RTT.

Right - noncompliant or nonstandard implementations can do various other
things.
(Continue reading)

Permalink | Reply | 142 comments |
headers
Ian McDonald | 3 Jan 2007 04:58
Picon
Favicon

Re: Are we doing sliding window in the Internet?

> > If you have ABC (as recent Linux senders do by default), or don't use
>
> ABC is EXPERIMENTAL.
>
And ABC is now off by default on even later kernels as basically the
congestion window didn't grow with how the whole code base interacted.

Can't comment on the delayed acks as don't know that part of the code so well.

Ian
--

-- 
Web: http://wand.net.nz/~iam4
Blog: http://imcdnzl.blogspot.com
WAND Network Research Group
Permalink | Reply | 3 comments |

Gmane