Re: [MEDIACTRL] [sip-overload] WGLC: draft-ietf-soc-overload-design

"Paul E. Jones" <paulej@packetizer.com> Tue, 31 August 2010 04:05 UTC

Return-Path: <paulej@packetizer.com>
X-Original-To: mediactrl@core3.amsl.com
Delivered-To: mediactrl@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 7543F3A68C7; Mon, 30 Aug 2010 21:05:36 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.267
X-Spam-Level:
X-Spam-Status: No, score=-1.267 tagged_above=-999 required=5 tests=[AWL=0.731, BAYES_00=-2.599, HTML_MESSAGE=0.001, J_CHICKENPOX_21=0.6]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id MEnzTugY-jbd; Mon, 30 Aug 2010 21:05:26 -0700 (PDT)
Received: from dublin.packetizer.com (dublin.packetizer.com [75.101.130.125]) by core3.amsl.com (Postfix) with ESMTP id AA9D63A67BE; Mon, 30 Aug 2010 21:05:25 -0700 (PDT)
Received: from sydney (rrcs-98-101-146-183.midsouth.biz.rr.com [98.101.146.183]) (authenticated bits=0) by dublin.packetizer.com (8.14.2/8.14.2) with ESMTP id o7V45gph006800 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NO); Tue, 31 Aug 2010 00:05:48 -0400
DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=packetizer.com; s=dublin; t=1283227548; bh=JWZIevDUHr/IeaDt5LKDDjuaxnFNiXIAyf3zGFjB42I=; h=From:To:Cc:References:In-Reply-To:Subject:Date:Message-ID: MIME-Version:Content-Type; b=cXjuzCwQ20JhwiSpF/euyimuEVBeABLdpxTO8mRgkYT7HPopyFYO8RU46kuICL25p Lzus+4gWPn4lUce+fAZ8Xrp8l2/s7Luke9YP+KQyKid5x9TIhcHy3mdJWN53Kxz4YZ mLbJLj7uIwwMeNR3dPsithuQNg4mvksYruY3hV68=
From: "Paul E. Jones" <paulej@packetizer.com>
To: "'Parthasarathi R \(partr\)'" <partr@cisco.com>, "'Volker Hilt'" <volker.hilt@alcatel-lucent.com>
References: <4C71B1C3.6070805@ericsson.com> <A11921905DA1564D9BCF64A6430A62390293A4AF@XMB-BGL-411.cisco.com><4C7AA34D.4020000@alcatel-lucent.com> <A11921905DA1564D9BCF64A6430A62390293A4B0@XMB-BGL-411.cisco.com> <4C7AC02D.1000200@alcatel-lucent.com> <OF5FC5A3A1.0A30DB2F-ON8525778E.006FC85F-8525778E.0070FB2C@csc.com> <A11921905DA1564D9BCF64A6430A623903054F93@XMB-BGL-411.cisco.com> <4C7BC713.3010208@alcatel-lucent.com> <A11921905DA1564D9BCF64A6430A62390293A4B6@XMB-BGL-411.cisco.com>
In-Reply-To: <A11921905DA1564D9BCF64A6430A62390293A4B6@XMB-BGL-411.cisco.com>
Date: Tue, 31 Aug 2010 00:04:18 -0400
Message-ID: <034e01cb48c1$9b406dd0$d1c14970$@packetizer.com>
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary="----=_NextPart_000_034F_01CB48A0.14313ED0"
X-Mailer: Microsoft Outlook 14.0
Thread-Index: AQF88xjIndSe1GjyNRVqStwhmMYsiALfdHr6AcNJVLcAzCvLDwHj6swUAwCiYBsBwkhZDQF4GhhkAPhdq4+TI2dJAA==
Content-Language: en-us
X-Mailman-Approved-At: Tue, 31 Aug 2010 08:04:13 -0700
Cc: sip-overload@ietf.org, mediactrl@ietf.org
Subject: Re: [MEDIACTRL] [sip-overload] WGLC: draft-ietf-soc-overload-design
X-BeenThere: mediactrl@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Media Control WG Discussion List <mediactrl.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/mediactrl>, <mailto:mediactrl-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/mediactrl>
List-Post: <mailto:mediactrl@ietf.org>
List-Help: <mailto:mediactrl-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/mediactrl>, <mailto:mediactrl-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 31 Aug 2010 04:05:36 -0000

Personally, I've viewed the work on SOC complementary to what we had also
proposed in the Session Capacity Estimate (SCE) draft.  I agree with Partha
here when he says that the current algorithm is only really focused on the
signaling plane and does not really consider other issues.

 

The fact is that a device that provides information to a peer suggesting a
reduction in the rate of messages is only looking at the high-level
transaction processing capacity.  One could, in theory, observe resources
and suggest an increase or reduction in the call signaling rate as resources
are reduced.  However, transaction processing capability is one thing, and
available memory, DS0s, DSPs, etc. is something else.

 

When there is proxy sitting in front of a bank of SBCs, something more is
needed than mere rate control.  The SBC needs to know which SBC likely has
the available resources to handle the call.  One could make this an
elaborate solution wherein the proxy knows tons of information about each
SBC, but the INVITE probably only provides a hint about the needs of the
call and, therefore, is having so much detail will not be very useful.  What
I think it useful, though, is knowing that SBC-1 is telling the proxy that,
given the current load it has, it can likely handle 10 more sessions, and
that SBC-2 is telling the proxy that, given the current load it had, it can
likely handle 100 more sessions.  Combining that SCE approach with a
rate-limiting mechanism provides us with a fairly complete solution.

 

Paul

 

From: sip-overload-bounces@ietf.org [mailto:sip-overload-bounces@ietf.org]
On Behalf Of Parthasarathi R (partr)
Sent: Monday, August 30, 2010 10:30 PM
To: Volker Hilt
Cc: Paul Jones (paulej); sip-overload@ietf.org; mediactrl@ietf.org
Subject: Re: [sip-overload] WGLC: draft-ietf-soc-overload-design

 

Volker,

 

In case you think that SIP overload control algorithm includes the other
loads in the signaling plane then the current algorithms mentioned in the
design document is not sufficient. In IETF-78 SoC interim meeting,
SIP-overload design experiments slides mentioned that media congestion is
not part of the experiment. 

 

SIP is the control protocol for RTP traffic. There is no need of separate
overload control for RTP in case SIP overload control algorithm is designed
for Media load as well. The access control mentioned by you has to be
considered as "local overload control" which is not suffice as per the
current SIP overload design as well as in the reality. 

 

Even I discussed about different RTP load in IETF-78 SoC as part of
draft-partha-soc-overload-resource-availability-00 presentation. I'll
explain again here:

 

B2BUA with address hiding service (SBC), memory requirements vary for
different types of call as indicated below

 

SIP Audio (RTP) call - 0.2%

SIP Video (RTP) call - 1%

SIP telepresence (RTP) call - 10%

 

The memory of SIP and RTP shall be shared across. RTP load varies depends
upon the negotiated codec during the call. The incoming SIP INVITE message
acceptance depends upon the dialog(call)-hold time as well as negotiated
codec. There is no relationship between the number of SIP messages and the
system load in case of SIP media servers. I'm sure that the above problem is
not addressed as part of the current design. Please explain me in case the
current algorithms solve the above mentioned problem.

 

I could not visualize that media congestion is not part of SIP overload
design (document) till I'm attending IETF-78 interim meeting. Please add the
explicit text in the SIP overload design document that Congestion collapse
in SIP server due to other load is outside scope of the document. The text
will surely help other implementers like me.

 

Thanks

Partha 

  _____  

From: Volker Hilt [mailto:volker.hilt@alcatel-lucent.com]
Sent: Mon 8/30/2010 8:28 PM
To: Parthasarathi R (partr)
Cc: Janet P Gunn; mediactrl@ietf.org; Paul Jones (paulej);
sip-overload@ietf.org
Subject: Re: [sip-overload] WGLC: draft-ietf-soc-overload-design

Partha,

if other load causes a congestion collapse on the SIP signaling plane of
a server, it is within the scope of an overload control algorithm.
Whenever a SIP server runs out of resources to process incoming SIP
messages, it needs to take action to make sure it is not encountering a
SIP congestion collapse.

It may use other mechanisms or intelligent design choices to control the
load generated by other processes on a SIP server. E.g., it may perform
access control for admitting audio streams for processing, use overload
control mechanism for other protocols, give less priority to other
processes (e.g., a virus scanner), etc.

Thanks,

Volker





On 8/30/2010 1:42 AM, Parthasarathi R (partr) wrote:
> My point is same as Janet's. I wish that SIP overload design document
> has to explicitly indicate about congestion collapse in SIP server due
> to other loads (like RTP message) is outside the scope of the document.
> Without this text in the document, the implementers of SIP servers with
> other load handling may be mislead.
> Thanks
> Partha
>
> ------------------------------------------------------------------------
> *From:* Janet P Gunn [mailto:jgunn6@csc.com]
> *Sent:* Monday, August 30, 2010 2:04 AM
> *To:* Volker Hilt
> *Cc:* mediactrl@ietf.org; Parthasarathi R (partr); Paul Jones (paulej);
> sip-overload@ietf.org; sip-overload-bounces@ietf.org
> *Subject:* Re: [sip-overload] WGLC: draft-ietf-soc-overload-design
>
>
> I guess the points are:
> 1- If a SIP actor is experiencing overload (regardless of whether the
> overload is due to "too many SIP messages" or something else entirely)
> it needs to reduce the rate at which SIP messages arrive at the SIP
> actor (the focus of this document).
> 2- If the overload is largely due to something other than SIP messages,
> then ONLY reducing the arrival of SIP messages may not significantly
> reduce the overload, and you may still have congestion collapse.
> 3- For devices which are SIP actors, and also serve other loads (e.g.
> RTP, database lookup, ISUP gateway), it is to be hoped that there is an
> overload control mechanism for each type of traffic, and that the
> mechanisms are coordinated in a sensible way (so that no one type of
> traffic is starved out by the others). But, other than a passing
> comment, that is out of scope for a document on SIP Overload Control.
>
> Janet
>
> This is a PRIVATE message. If you are not the intended recipient, please
> delete without copying and kindly advise us by e-mail of the mistake in
> delivery.
> NOTE: Regardless of content, this e-mail shall not operate to bind CSC
> to any order or other contract unless pursuant to explicit written
> agreement or government initiative expressly permitting the use of
> e-mail for such purpose.
>
>
> From:         Volker Hilt <volker.hilt@alcatel-lucent.com>
> To:   "Parthasarathi R (partr)" <partr@cisco.com>
> Cc:   "mediactrl@ietf.org" <mediactrl@ietf.org>rg>, "sip-overload@ietf.org"
> <sip-overload@ietf.org>rg>, "Paul Jones \(paulej\)" <paulej@cisco.com>
> Date:         08/29/2010 04:16 PM
> Subject:      Re: [sip-overload] WGLC: draft-ietf-soc-overload-design
>
>
> ------------------------------------------------------------------------
>
>
>
> Partha,
>
> the CPU load of a SIP server can increase for many reasons, not only the
> processing of RTP packets. For example, the system might start the daily
> backup of a database, apply patches, run a virus scan or launch another
> application.
>
> If any of these events leads to a shortage of CPU cycles in which the
> SIP server may not be able to process all incoming SIP messages, it
> should use overload control to reduce the number of SIP messages.
>
> Thanks,
>
> Volker
>
>
>
>
> On 8/29/2010 4:06 PM, Parthasarathi R (partr) wrote:
>  > Volker,
>  > I doubt whether the current draft will serve for all SIP servers. I'm
>  > taking the example of B2BUA with media (RTP)topology hiding
>  > functionality and there is no DSP, DS0 or DISK resource involved. Here,
>  > Server experiences an overload of SIP message not just because of
>  > incoming SIP messages but also due to the incoming RTP messages and RTP
>  > messages flow varies depend upon the dialog (call) hold time whereas
>  > there may not any SIP message when the dialog is established. SIP
>  > processor mentioned in the system model of this draft may be shared the
>  > physical processor with RTP messages or other protocols messages. We
>  > have the problem in incorporating the current SIP overload design in
>  > media handling SIP devices.
>  > I'm including mediactrl to hear their experience with SIP overload
>  > design and mediactrl combined SIP implementation.
>  >
>  > Thanks
>  > Partha
>  >
------------------------------------------------------------------------
>  > *From:* Volker Hilt [mailto:volker.hilt@alcatel-lucent.com]
>  > *Sent:* Sun 8/29/2010 11:43 PM
>  > *To:* Parthasarathi R (partr)
>  > *Cc:* Salvatore Loreto; sip-overload@ietf.org
>  > *Subject:* Re: [sip-overload] WGLC: draft-ietf-soc-overload-design
>  >
>  > Partha,
>  >
>  > the design document discusses problems related to the overload of SIP
>  > messages. Any server that processes SIP messages can experience an
>  > overload of SIP messages. Therefore, the discussions in the draft apply
>  > to all servers that are involved in SIP message processing.
>  >
>  > SIP servers that also provide media handling services may need
>  > mechanisms to signal that they are out of trunk lines, DSP processors,
>  > storage capacity for voicmails, etc. These mechanisms will often be
>  > needed at times when the servers can still process SIP messages. These
>  > mechanisms are out of scope for the working group as discussed.
>  >
>  > Thanks,
>  >
>  > Volker
>  >
>  >
>  >
>  >
>  > On 8/29/2010 1:21 PM, Parthasarathi R (partr) wrote:
>  > > Sal/Volker,
>  > > The design document is designed for dedicated SIP signaling servers
and
>  > > experiments are done based on the dedicated servers. The same
> design may
>  > > not scale for SIP server with media handling. As per IETF-78 meeting,
>  > > SIP servers with media handling
>  > > (draft-partha-soc-overload-resource-availability-00,
>  > > draft-jones-sip-overload-sce-00) is outside the scope of this working
>  > > group. I think that it is better to indicate the scope of this
overload
>  > > design is for "dedicated SIP servers" only and does not include media
>  > > servers with SIP. Please let me know your opinion on the same.
>  > > Thanks
>  > > Partha
>  > >
> ------------------------------------------------------------------------
>  > > *From:* sip-overload-bounces@ietf.org on behalf of Salvatore Loreto
>  > > *Sent:* Mon 8/23/2010 4:54 AM
>  > > *To:* sip-overload@ietf.org
>  > > *Cc:* Volker Hilt
>  > > *Subject:* [sip-overload] WGLC: draft-ietf-soc-overload-design
>  > >
>  > > [as chair]
>  > >
>  > > the draft has been update to address all the concerns/suggestions
that
>  > > have been raised during
>  > > the F2F meeting in Maastricht:
>  > >
>  > > http://tools.ietf.org/html/draft-ietf-soc-overload-design-01
>  > >
>  > >
>  > > the chairs believe the document has no remaining open issues, and is
>  > > ready for WG last call evaluation.
>  > >
>  > > Today we are starting a two-weeks working group last call.
>  > > This call end on September 5th.
>  > >
>  > > comments and especially full and deep reviews of the draft are really
>  > > appreciate.
>  > >
>  > >
>  > > cheers
>  > > Sal
>  > >
>  > > --
>  > > Salvatore Loreto
>  > > www.sloreto.com
>  > >
>  > > _______________________________________________
>  > > sip-overload mailing list
>  > > sip-overload@ietf.org
>  > > https://www.ietf.org/mailman/listinfo/sip-overload
>  > >
>  >
> _______________________________________________
> sip-overload mailing list
> sip-overload@ietf.org
> https://www.ietf.org/mailman/listinfo/sip-overload
>
>