|
Hi all, We have updated the GMPLS LSP DPPM draft recently (see announcement below).
The draft has received positive responses after the last CCAMP meeting. A
discussion group was later setup by Guoying Zhang (RITT, MII of China). Thanks
for those who has participated in the group and gave suggestions. In particular,
we would like to thank Monique Morrow (Cisco) for indicating a new draft in
benchmarking WG related to RSVP-TE performance authored by Cisco folks (http://www.ietf.org/internet-drafts/draft-vapiwala-bmwg-rsvpte-convergence-motivation-00.txt).
We regard this as interests in GMPLS performance, whose progress will help
motivating and progressing the GMPLS LSP DPPM draft. The major improvements in this version include: 1) We modified the intro part to state the fact that the metrics defined
are also applicable for TE-MIB inclusion, as may be interest of carriers to monitor
the control plane performance in realtime, through NMS. 2) We corrected some error pointed out by X. Wei(Fiberhome) and J. Gao(Huawei). 3) We changed the measurement unit from "seconds" to
"milli-seconds" as devices that have an averaged per-node processing
delay of less than 100ms is readily available. During preparation of this version, there has been controversy on whether
and how LSP rerouting metrics should be added to the draft. Carriers have shown
their interest in this metric. However, it seems to us that the measurement of rerouting
time will inevitably change the simplicity of methodologies that have been
introduced in the draft so far, and the measurement may also be tightly coupled
with implementations. Finally Guoying and I would rather leave it to the
working group for possible suggestions and solutions. Please kindly give your comments and suggestions. Regards, Weiqiang Sun To: i-d-announce at
ietf.org Subject: I-D
ACTION:draft-xie-ccamp-lsp-dppm-01.txt From: Internet-Drafts at
ietf.org Date: Wed, 30 May 2007
18:50:01 -0400 A New Internet-Draft is
available from the on-line Internet-Drafts directories. Title :
Label Switched Path (LSP) Dynamical Provisioning Performance Metrics in
Generalized MPLS Networks Author(s) :
G. Xie, et al. Filename :
draft-xie-ccamp-lsp-dppm-01.txt Pages :
34 Date :
…A URL for this Internet-Draft is:
-----Original Message----- I shall do this -- stay tuned! On 5/25/07 9:38 PM, "Weiqiang Sun" <sunwq@sjtu.edu.cn>
wrote: > Hi Monique, > > It's nice that we have one other language in common :) > > Last month when Fred Baker visited SJTU, we were given five minutes
to > introduce the draft. Mr. Baker showed interest in this and we also > mentioned your involvement in this :). > > We are happy to see that cisco folks also started to work on this, > though in BMWG. It would be great if you could kindly introduce the > authors of the draft into this discussion group and possiblly
arrange a > discussion in the next ietf meeting for us, which I believe will > increase mutual understanding and help fasten the standardization > process of metrics and related methodologies. > > Thanks. > > -- > Weiqiang > SJTU > > > > On 2007-05-25 08:06 -0700，Monique Morrow
Wrote: > > Nimen hao! >> >> Your timing is perfect as I was about to ping you on the draft. >> >> You may want to look at new draft for possible harmonization
with the GMPLS >> LSP DDPM draft: >> >>
http://www.ietf.org/internet-drafts/draft-vapiwala-bmwg-rsvpte-convergence-m >> otivation-00.txt >> >> >> Kind regards, >> >> Monique >> >> On 5/25/07 1:35 AM, "Weiqiang Sun"
<sunwq@sjtu.edu.cn> wrote: >> >>> Hi all, >>> >>> We have just updated the draft GMPLS LSP DPPM according the
feedbacks we get >>> since its publication on IETF website. Many thanks to those
who have given >>> valuable suggestions and also to Adrian and Deborah for
allocating the slot >>> in the last meeting. >>> >>> The major improvements are: >>> 1) We modified the intro part to state the fact that the
metrics defined are >>> also applicable for TE-MIB inclusion, as may be interest of
carriers to >>> monitor the control plane performance in realtime, through
NMS. >>> 2) We corrected some error pointed out by X. Wei(Fiberhome)
and J. >>> Gao(Huawei). >>> 3) We changed the measurement unit from "seconds"
to "milli-seconds" as >>> devices that have an averaged per-node processing delay of
less than 100ms >>> is readily available. >>> >>> The updated draft is attached for your review, before we
submit it to IETF >>> for publication. Please give your comments. >>> >>> Thanks and regards. >>> >>> Weiqiang >>> SJTU >>> -----Original Message----- >>> From: Weiqiang Sun [mailto:sunwq@sjtu.edu.cn] >>> Sent: Thursday, April 05, 2007 1:11 PM >>> To: 'BRUNGARD, DEBORAH A, ATTLABS'; >>> Jianghui'; >>> >>> Cc: 'Weisheng HU'; 'jinyh@sjtu.edu.cn'; 'zhaohd91@163.com' >>> Subject: RE: New discussion group formed on draft GMPLS LSP
DPPM >>> >>> Hi all, >>> >>> It is really nice to have all of you together to discuss
this draft. The >>> draft team has had a difficult time when the draft was in
its primary stage >>> months ago. Now I believe your valuable suggestions and
comments will make >>> it a lot easier. >>> >>> As you may be aware, the draft deals only with control
plane performance, in >>> terms of circuit provisioning capability. One reason that
justifies this >>> work, IMHO, is that the control plane not only automates
the provisioning >>> process, but more importantly, it also significantly
reduces the setup >>> delay. This opens up opportunities for applications or
services that rely on >>> dynamically provisioned circuits. One such application has
been implemented >>> in a GMPLS testbed in >>> systematic way to evaluate the provisioning performance of
the control >>> plane, especially when the circuit hold-time is comparable
with or several >>> times the provisioning delay. A set of standardized metrics
that vividly >>> characterize the performance of control plane will
definitely increase the >>> confidence of applying GMPLS networks in the most natural
way and help to >>> design a proper timing model. >>> At the same time, as you may have seen in Guoying's talk
during the last WG >>> meeting, although variations are observed for all traffic
loads >>> (dynamicity), higher load will lead to higher variation in
provisioning >>> delay. As control plane scalability and resilience issues
are getting into >>> our sight, a set of standardized metrics will hopefully
stimulate the >>> improvement of control plane implementations, e.g., using
hardware >>> acceleration to reduce the processing time and variation. A
reliable and >>> stable control plane implementation will in turn encourage
novel >>> applications being devised. >>> Hope Monique would be satisfied with this in regarding to
question 2. :) >>> >>> As to the metrics themselves, the existing ones are merely
about control >>> plane performance, with no regard to CP DP synchronization.
This is the most >>> lightweight solution we can think of so far and is
reasonable when the time >>> needed to establish a local XC is neglectable if compared
with signaling >>> processing delay. If this is not the case, still, we can
use the control >>> plane delay, plus a constant representing the XC operation
overhead. >>> By the term provisioning, we have intended to describe
application models >>> that circuits are established PRIOR to data transfer. This
is different from >>> the reroute model, where data transfer are ongoing and the
rerouting >>> procedure will eventually leads to packet loss (or service
interruption). >>> Thus a more reasonable metric that describes rerouting
performance, in my >>> opinion, would be packet loss, or service interruption
time, not delay (in >>> the control plane's point of view).It seems to me that
measurement of GMPLS >>> LSP rerouting is very similar to measurement of MPLS
protection, which is >>> now been defined in BMWG. Personally, I suggest we initiate
another draft to >>> deal of LSP rerouting, with reference to MPLS protection in
BMWG. Hopefully >>> we can get more suggestions from Al in this direction. >>> >>> For the first question raised by Monique, I have nothing to
say by now since >>> we are still working on the mentioned documents. Should you
have any >>> comments, please let us know. >>> >>> For the discussion group, we have a new member, Dr. Huandong
Zhao from >>> Telecom Shanghai Research Institute. >>> >>> Cheers, >>> Weiqiang >>> SJTU >>> -----Original Message----- >>> From: BRUNGARD, DEBORAH A, ATTLABS
[mailto:dbrungard@att.com] >>> Sent: Wednesday, April 04, 2007 1:56 AM >>> To: Monique Morrow; zhangguoying; Weiqiang Sun; Han
Jianghui; blithe; xqwei; >>> gjhhit; lufang; i-nishioka; julien.meuric; MORTON, ALFRED
C, JR. ( >>> ATTLABS >>> Subject: RE: New discussion group formed on draft GMPLS LSP
DPPM >>> >>> Hi all, >>> >>> I've added Al's address- >>> Deborah >>> >>> -----Original Message----- >>> From: Monique Morrow [mailto:mmorrow@cisco.com] >>> Sent: Monday, April 02, 2007 5:16 AM >>> To: zhangguoying; Weiqiang Sun; >>> BRUNGARD, DEBORAH A, ATTLABS; lufang; i-nishioka;
julien.meuric >>> Subject: Re: New discussion group formed on draft GMPLS LSP
DPPM >>> >>> Greetings colleagues, >>> >>> This is an excellent proposal in terms of focused dialogue
with regard to >>> this topic. >>> >>> To progress: >>> >>> >>> - Important is note what work has been done in this
direction <Al Morton >>> has done quite a bit of work -- in this direction for
example in the ITU >>> <y.1540, y.1541, y.1542> -- gap analysis as to what
is missing; >>> >>> - Agreement as to the metrics themselves ; >>> >>> - Problem to be solved -- why is this an important ?
<somewhat implied in >>> the draft; >>> >>> - Method as to actual mechanism to measure would be last
IMHO; >>> >>> >>> Looking forward - >>> >>> /m >>> >>> >>> >>> >>> On 2/4/07 12:08 am, "zhangguoying"
<zhangguoying@mail.ritt.com.cn> wrote: >>> >>>> Hi all, >>>> >>>> Fist of all, I suggest we form a new discussion group
to include experts >>> who >>>> are interested in the "GMPLS LSP dynamic
provisioning performance draft" >>>> during the IETF 68 meeting. >>>> Let me say welcome to: >>>> >>>> Deborah A. Brungard, AT&T, >>>> Monique Jeanne Morrow, Cisco, >>>> Luyuan Fang, Cisco >>>> Itaru Nishioka, NEC >>>> >>>> Deborah would also introduce her colleague Al Morton,
chair of the >>>> benchmarking methodology WG to join our discussion. >>>> >>>> The previous discussion group include : >>>> Guowu Xie,SJTU >>>> Weiqiang Sun,SJTU >>>> Guoying Zhang, CATR MII >>>> Jianghui Han, IXIA >>>> Xueqing Wei, Fiberhome >>>> Jianhua Gao, Huawei >>>> >>>> >>>> Here are some important suggestions I received during
the meeting: >>>> 1. Hear more voices and comments from the carriers.
Currently the draft >>> could >>>> be discussed in ccamp group, but may be transfered to
the management AD >>>> later.---- >>>> 2. Refer to the work of benchmarking WG. ---- Deborah >>>> 3. Include the performance metrics of LSP rerouting.
These performance >>> metrics >>>> would be useful if the GMPLS MIB support monitoring
them in deployed >>>> network.---- Julien >>>> 4. Define the method on how to measure the performance
metrics, and refer >>> to >>>> the work of benchmarking WG. ----Itaru Nishioka >>>> >>>> >>>> Next we should discuss the plan for 01 version. I
suggest we fist refer to >>> the >>>> benchmarking methodology WG work, and consider the
possibility of >>> including >>>> the LSP rerouting performance in the 01 version. >>>> >>>> >>>> >>>> zhangguoying >>>>
>>>>
>>>> zhangguoying@mail.ritt.com.cn >>>> |