SCTP Discussions

Text archives Help


Re: AW: [SCTP-Discussion] Re: Bundling


Chronological Thread 
  • From: Michael Tuexen <Michael.Tuexen AT micmac.franken.de>
  • To: discussion AT sctp.de
  • Subject: Re: AW: [SCTP-Discussion] Re: Bundling
  • Date: Mon Mar 18 16:12:01 2002
  • List-id: SCTP Discussions <discussion.sctp.de>

Dear all,

this is not a trivial API! I'm not sure if this complexity is what the
user wants.

In my opinion the user might want to tell the SCTP stack: Please do not
delay the sending only to have a chance for bundling.

So my understanding of the constants is:
SCTP_BUNDLING_DISABLED: Send as soon as possible, do not wait for further DATA to bundle.
SCTP_BUNDLING_ENABLED: Waiting a definite time (has to be provided with a different
mechanism) is allowed.

That was my understanding of the API, but we have not implemented the timer based delay.
That could be a feature for a future release but I think other features are more important.

Best regards
Michael

On Monday, March 18, 2002, at 02:18 PM, Schruefer Wolfgang wrote:

Hello Andreas,

you asked for suggestions...

Here is a suggestion for an API to give the upper layer nearly full contol for bundling.
It is realised with the control-flag 'dontBundle' in sctp_send() and an additional Instance- (or Association-) Parameter 'bundlingSendDelay' to define the bundling time.

#define SCTP_BUNDLING_ENABLED 0
/* Like SCTP_SEND_IMMEDIATELY (for historic reasons)*/

#define SCTP_BUNDLING_DISABLED 1
/* Like SCTP_SEND_IMMEDIATELY (for historic reasons)*/

#define SCTP_SEND_IMMEDIATELY 2
/* The Packet is created procedural and sent out if allowed by receiver*/

#define SCTP_SEND_DELAYED 3
/* A timer (duration=InstanceParameter: bundlingSendDelay, 0 allowed!) is started (if not running yet).
The Packet is sent out at timer expiry or if maximum packet size was reached */

#define SCTP_DONT_SEND_NOW 4
/* The message is only queued in the send queue, sctp needs external trigger for sending.
Exception: it will be sent out immediately if the maximum packet size was reached*/

#define SCTP_SEND_NOW 5
/* Trigger to send out queued messages. A call of sctp_send() with this flag may contain
no message attached (then only Params: *buffer==NULL, associationID) */


Viele Gruesse
Wolfgang


-----Ursprüngliche Nachricht-----
Von: Andreas Jungmaier
[mailto:ajung AT exp-math.uni-essen.de]
Gesendet: Montag, 18. März 2002 10:44
An:
discussion AT sctp.de;
Alberto Bellettato
Cc: Michael Tüxen
Betreff: Re: [SCTP-Discussion] Re: Bundling


Alberto,

bundling generally implies that the implementation wait some
time before data is sent. If, during that time another chunk
is sent by the application, it may be bundled.
If, however, the application does not send any data within that
time, no bundling is used, but performance may be worse, because
of the additional delay.

Now, actually that is a matter of the API.

And this is a question to the list: are there any suggestions/
comments as to how that might be handled ?

Regards,

Andreas


On Saturday, 16. March 2002 10:34, Alberto Bellettato wrote:
I am testing the multi-streaming feature and I think that the absence of
head-of-line blocking within streams should give better performances in
case of bundling utilization (but only if chunks with _different_ stream ID
are bundled inside of the same SCTP packet ).
For this reason I need to understand exacly when bundling is used
(unfortunately I could not extrapolate the implementation's logic from
Ethereal logs).
For this multi-streaming test, I would need that bundling is always used to
get top performances. But, setting SCTP_BUNDLING_ENABLED, bundling is _not_
always used. Then, under what conditions is it not used?

Best Regards,
Alberto

From: Tuexen Michael
<Michael.Tuexen AT icn.siemens.de>
To:
"'discussion AT sctp.de'"

<discussion AT sctp.de>
Cc: "'Alberto Bellettato'"
<alberto.bellettato AT icn.siemens.it>
Subject: RE: [SCTP-Discussion] Bundling
Date: Thu, 14 Mar 2002 15:23:09 +0100
Reply-To:
discussion AT sctp.de

Hi Alberto,

we provide to constants:
SCTP_BUNDLING_ENABLED
SCTP_BUNDLING_DISABLED

These affect the initial sending of there is not flowcontrol
or congestioncontrol active which delays the sending. In case
of retransmissions or delayed sending bundling can occur if
it results in better protocol performance.
Even if you set SCTP_BUNDLING_DISABLED the DATA chunk may
be bundled on retransmissions.

Why do you need this behaviour for testing? It should only be
avoided that SCTP restricts the protocol performance unnecessary
by waiting for data for bundling.

Best regards
Michael

_______________________________________________
discussion mailing list
discussion AT sctp.de
http://www.sctp.de/mailman/listinfo/discussion

--
------------------------------------------------------------------------
Dipl.-Ing. Andreas Jungmaier | _
Computer Networking Technology Group | ASCII ribbon campaign ( )
University of Essen | - against HTML email X
http://www.exp-math.uni-essen.de/~ajung | & vcards / \_______________________________________________
discussion mailing list
discussion AT sctp.de
http://www.sctp.de/mailman/listinfo/discussion
_______________________________________________
discussion mailing list
discussion AT sctp.de
http://www.sctp.de/mailman/listinfo/discussion


Michael.Tuexen AT micmac.franken.de





Archive powered by MHonArc 2.6.19+.

Top of page