The MPLS WG Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] LSP setup time and delay
Thanks for the clarifications:
In the MPLS network, the LSRs exchange the
OSPF TE LSAs containing bandwidth availability
information etc.
Now if the resource avail changes and it calls for a
new TE LSP, would this change (i.e. new LSP
establishment) happen automatically or would it be
data-driven.
i.e. if the OSPF Routing Plane gets a new TE
LSA, it could handle it in following ways:
a. Routing plane runs the SPF calculation
on receipt of every TE LSA and informs the MPLS
Signalling Plane of this change.
If we follow (a), it could mean:
- more SPF calculations and more computation
overhead on the LSRs.
b. Routing Plane waits for the Signaling
Plane to trigger a request for a new explicit
path. Once requested, Routing Plane runs the
SPF algo, calculates new routes and gives it
back to the Signalling Plane. i.e. the SPF
calculation is data-driven.
If we follow (b), it could lead to
- delays (as mentioned in the original ques).
Which of these is preferred approach and why ?
As far as I remember reading on this, approach (b)
is preferred. But, if that is correct, it contradicts
with our current discussion (that LSP setup is _not_
data-driven).
Am I missing something...
Could you please throw some pointers...
Thanks,
Vishal.
--- Shahram Davari <Shahram_Davari@pmc-sierra.com>
wrote:
> Hi Vishal,
>
> >-----Original Message-----
> >From: Vishal M [mailto:vishal_study@yahoo.com]
> >Sent: Friday, August 18, 2000 2:11 PM
> >To: Shahram Davari; 'Tim.Cook@go.ecitele.com'
> >Cc: 'Vishal M'; mpls@UU.NET
> >Subject: RE: LSP setup time and delay
> >
> >
> >
> >Hi Shahram,
> >
> >Thanks for reply. Some more doubts:
> >
> >1. If the LSP setup is not data-driven, how
> > many LSPs are established on the ingress LSR
> (LER)
> > when LER initializes ?
>
> It depends on the network policy and requests that
> are signaled. In many
> cases (such as in a Diffserv network) you probably
> need at least one LSP per
> egress LSR.
>
> >
> >2. Wouldn't this lead to a lot of manual
> intervention
> > in pre-establishing these LSPs ?
>
> Not really. You just need to specify the ingress
> LSR, the egress LSR, the
> FEC and the required QoS (such as BW) for the
> required LSP, and the TE
> software will find and establish the path for you.
>
> >
> >3. What happens if LER gets an IP packet for which
> > there is no pre-established LSP (and assuming
> > that the IP packet belongs to a mission-critical
> > application) ?
> > How do we ensure QoS in this case ?
>
> If after the packet classification it is found that
> the packet does not map
> to any LSP, then this is up to the network policy,
> it may send it as L3
> packet or may drop it. It is up to the network
> provider to assign all
> required FECs to proper LSPs at the edge.
>
> The best method is for the mission-critical
> application to signal its
> required QoS. The network will use this information
> to establish the
> required LSP. If such an LSP can't be found then the
> request will be
> rejected. In other words admission control is the
> best method to make sure
> that the mission-critical packets receive their
> required QoS.
>
> Regards,
> -Shahram
>
> >
> >Thanks,
> >Vishal.
> >
> >--- Shahram Davari <Shahram_Davari@pmc-sierra.com>
> >wrote:
> >> Tim,
> >>
> >> No. Control packets are L3 forwarded. Also in
> data
> >> driven method, the
> >> packets are initially L3 forwarded and then
> switched
> >> to L2 forwarding when
> >> LSP is setup.
> >>
> >> -Shahram
> >>
> >> >-----Original Message-----
> >> >From: Tim.Cook@go.ecitele.com
> >> [mailto:Tim.Cook@go.ecitele.com]
> >> >Sent: Friday, August 18, 2000 10:49 AM
> >> >To: Shahram Davari
> >> >Cc: 'Vishal M'; mpls@UU.NET
> >> >Subject: RE: LSP setup time and delay
> >> >
> >> >
> >> >
> >> >
> >> >Does this mean that there is no L3 forwarding
> >> between two LSRs
> >> >(all traffic
> >> >is labeled)?
> >> >
> >> >
> >> >
> >> >
> >> >Shahram Davari <Shahram_Davari@pmc-sierra.com>
> on
> >> 08/18/2000
> >> >10:22:40 AM
> >> >
>
> >>
> >> >
>
> >>
> >> >
>
> >>
> >> > To: "'Vishal M'" <vishal_study@yahoo.com>,
> >> mpls@UU.NET
> >> >
>
> >>
> >> > cc: (bcc: Tim Cook/Fort Lauderdale/ECI
> >> Telecom)
> >> >
>
> >>
> >> >
>
> >>
> >> >
>
> >>
> >> > Subject: RE: LSP setup time and delay
>
> >>
> >> >
>
> >>
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >Vishal,
> >> >
> >> >What you are describing is called "data driven
> >> label
> >> >assignment". This is
> >> >explained in framework document, and requires L3
> >> forwarding
> >> >until the LSP
> >> >is
> >> >setup. Due to its problems it has not been
> pursued
> >> further by MPLS WG
> >> >(however, some implementations exist). Control
> >> driven label
> >> >assignment is
> >> >the one most of the people are working on.
> >> >
> >> >Regards,
> >> >-Shahram
> >> >
> >> >>-----Original Message-----
> >> >>From: Vishal M [mailto:vishal_study@yahoo.com]
> >> >>Sent: Friday, August 18, 2000 4:15 AM
> >> >>To: mpls@UU.NET
> >> >>Subject: LSP setup time and delay
> >> >>
> >> >>
> >> >>
> >> >>Hello,
> >> >>
> >> >>Suppose an IP packet is to sent over a MPLS
> >> network.
> >> >>As I understand the protocol, different
> components
> >> at
> >> >>the ingres LSR (also called LER) interact in
> the
> >> >>following way:
> >> >>
> >> >>1. MPLS Signaling Plane (CR-LDP/RSVP) queries
> the
> >> >>Routing Plane (IGP Extensions) to get the
> Explicit
> >> >>Path to reach the destination (assuming nothing
> >> >>in the tables initially).
> >> >>
> >> >>2. Signaling Plane uses RSVP or CR-LDP to
> reserve
> >> >>resources along the explicit path. Label
> mapping
> >> >>is performed at the intermediatory LSRs.
> >> >>
> >> >>3. After the path is set up i.e. LSP is
> >> established
> >> >>from the source to the destination, the
> incoming
>
=== message truncated ===
__________________________________________________
Do You Yahoo!?
Send instant messages & get email alerts with Yahoo! Messenger.
http://im.yahoo.com/
|
|