Re: [mpls-linux-general] Re: MPLS forwarding /IP forwarding

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi James

Since the limitation of hardware, I don't have many PC and 1000M fast
ethernet card, But I still want emulate the large network

I try emulate the large IP routing table in core LSR . the entries of my IP
routing table is 5000. I expected to see the latency with IP increase, but
MPLS no change,,

Unfortuanly expermient show still no diffserence with or withour large IP
table.

is there some way to emulte the large MPLS network?

thanks

yuxiao

----- Original Message -----
From: "James R. Leu" <jl...@mi...>
To: "yuxiao jia" <yx...@si...>
Cc: <mpl...@li...>
Sent: Tuesday, June 17, 2003 7:22 PM
Subject: Re: [mpls-linux-general] Re: MPLS forwarding /IP forwarding

> You will be hard pressed to notice latency variation in such a small
> network, no matter what protocol you use.  Try measuring throughput
> instead.  Also, as I said in my previous e-mail, as soon as you introduce
> a LER (ingress or egress) into your throughput test, you will see a
> decrease in performance compared to IP only.  For you to see the benifits
> or MPLS forwarding (with respect to throughput measured on the LSR).  You
> will have to have many LERs connected to your LSR(s).
>
> For example:
>
> iL = ingress LER
>
>              100Mbps                 1000Mbps
>
>        iL1-----.
>                 \
>        iL2-----. \     ------------            -----------
>        .        \ `---|            |          |           |
>        .         `----|   LSR 1    |----------|  LSR 2    |
>        .         .----|            |          |           |
>        .        / .---|            |          |           |
>        iL(N-1)-. /     ------------            -----------
>                 /
>        iL(N)---`
>
> In this case if each ingress LER generated 100Mbps of labeled traffic
> you should see N*100 Mbps of traffic on the core link (where N is < 10).
> Typically IP forwarding on Linux will top out at about 600Mbp (depending
> on the packet size).
>
> On Tue, Jun 17, 2003 at 05:09:40PM -0400, yuxiao jia wrote:
> > hello James
> >
> > I measure per node delay in MPLS network (and corresponding IP network).
> > Toplolgy:  Ingress- Core---Egress
> >
> > The way I did is use external PC with two interafce connect to router ,
> > running tcpdump for capure IP forwarding, runing ethereal/tethereal to
> > caputer MPLS packet.
> >
> > ------router -------
> >     |               |
> >       analyser
> >
> > based on same packet ID, I calculate the differenc between incoming
timstamp
> > and outgoing timestamp. I believe the per node delay what I got is =
> > forarding +queusing delay+ transimiiton delay
> >
> >
> >
> > ----- Original Message -----
> > From: "James R. Leu" <jl...@mi...>
> > To: "yuxiao jia" <yx...@si...>
> > Sent: Tuesday, June 17, 2003 6:43 PM
> > Subject: Re: MPLS forwarding /IP forwarding
> >
> >
> > > The MPLS forwarding mechanism in mpls-linux is _not_ based on IP
> > forwarding.
> > > I believe Pim was refering to ingress and egress LER functionallity.
> > > In those cases it is expect that the forwarding will be slower then
> > > just straight IP because both IP and MPLS lookups need to occur (this
> > > case, but it hasn't).
> >
> > what does you meane "can be optimized in the egress LER  case"?
> >
> >
> > > As far as LSR forwarding being not being faster then IP, I would have
to
> > > look at your methodology for testing.  Are you measuring only latency?
> > > Or total throughput?  It has been a while since I've done any sort of
> > > erformance comparison, put last time performace testing was done, LSR
> > > forwarding was able to achive more throughput them IP forwarding.
> > I measure latency, how to measure total throuhput? can y give me some
ideas
> >
> > > In addition, the existance of mpls-linux is _not_ to provide a faster
> > > means of forwarding packets, it is meant to be a tool for people to
learn
> > > about MPLS.  To that end, if you really want to understand MPLS
> > > forwarding, look at the code.
> > >
> > > On Tue, Jun 17, 2003 at 04:35:03PM -0400, yuxiao jia wrote:
> > > > Hello Jame
> > >
> > > James not Jame.
> > >
> > > >
> > > > I am working the Linux testbed based on "RSVP-TE for MPLS diffserv"
> > > >
> > > > I want compare MPLS forwarding and normal IP forwarding perfomace.
The
> > topology is
> > > >
> > > >
> > > > -----Ingress -----Core ----10M -----Egress----  (other part is 100M)
> > > >
> > > > I measure the each router per node delay by using extrenal anlyaer.
the
> > result show
> > > >
> > > > on Ingress, since MPLS stack over head, the delay of MPLS > IP
> > > > on Core,  delay of MPLS almost = delay of IP ( I expected the MPLS
> > should fast then IP, since label swictc)
> > > > on Egress, delay of MPLS a liitle big , almost same =IP..
> > > >
> > > > I try figure out why on Core mpls forwarding not fast, I email Pim,
he
> > said under linux , MPLS forwarding is based on IP forwarding (since time
> > cosuming and error prone if withou IP forwarding)
> > > >
> > > > This is why I want ask you question
> > > >
> > > > 1) why in your mpls implemntaio, based on IP forawrding, does it do
IP
> > table look up?
> > > > 2) How much involved in IP forwarding when implement MPLS
forwarding.?
> > > > 3) if I emulete a lager IP routing table, can i get mpls fast than
IP?
> > > >
> > > > thanks
> > > >
> > > > yuxiao
> > > >
> > > >
> > >
> > > --
> > > James R. Leu
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: INetU
> > Attention Web Developers & Consultants: Become An INetU Hosting Partner.
> > Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
> > INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php
> > _______________________________________________
> > mpls-linux-general mailing list
> > mpl...@li...
> > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
>
> --
> James R. Leu
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: INetU
> Attention Web Developers & Consultants: Become An INetU Hosting Partner.
> Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
> INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general