The MPLS-OPS Archive

Cell Relay Retreat>MPLS-OPS Archive>month:2002-Dec> msg00116



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: LDP Hello and Keepalive intervals/holdtime

  • From: Carlos Patriawan <carlos@xebeo.com>
  • Date: Wed, 11 Dec 2002 17:40:26 -0500
  • CC: mpls-ops@mplsrc.com
  • Organization: Xebeo Communications
  • Resent-Date: Wed, 11 Dec 2002 19:15:03 -0500
  • To: Vishal M <vishal_study@yahoo.com>

from rfc 3036, sec 2.5.6-> when LSR rcv any LDP PDU messages (can be any
msg
other than hello/Keepalive), LSR must restart  its session Holdtimer, so
in below
case session wouldn't be terminated because these LDP/TCP(eg:Label
Mapping msg )
are treated "like" an hello/KA.

Carlos

Vishal M wrote:

> Hi Carlos,
>
> I agree that LDP applications could bundle multiple
> LDP mappings into a single big-TCP buffer to save on
> BW.
>
> The issue I was trying to state was of
> routing/topology change affecting multiple LDP
> sessions. For the duration this (routing change
> resulting in lots of LDP messages) is happening, the
> default Hello/KA holdtimers value may be a bit low.
>
> Has anyone seen this kind of behavior?
>
> Regards,
> Vishal.
>
> --- Carlos Patriawan <carlos@xebeo.com> wrote:
> > Vishal M wrote:
> >
> > > Hi,
> > >
> > > According to RFC 3036, LDP Hello interval is 5
> > sec, KA
> > > is 15 sec, Hello Holdtime is 15 sec and KA
> > Holdtime is
> > > 40 sec.
> > >
> > > If a LSR is acting in DU mode and has to deal with
> > > large number of FECs (say 100k), the holdTimes of
> > 15
> > > (hello) and 40 (KA) is quite small and may result
> > in
> > > timeouts when LDP (and/or IP) modules of the LSR
> > are
> > > busy transmitting/receiving these large number of
> > TCP
> > > messages (i.e. FEC mappings).
> > >
> > > Hence the operator may have to tune these
> > holdtimers
> > > to a higher number.
> > >
> > > Has anyone seen this kind of issue and would like
> > to
> > > share the experiences.
> > >
> > > What is a *realistic* (or practical) value for
> > > Hello/KA holdtimers (say given a gigabitEthernet
> > kind
> > > of interface) in ISP world ?
> > >
> > > Could someone please comment on this.
> > >
> > > Thanks for your time,
> > >
> > > Regards,
> > > Vishal.
> > >
> > > __________________________________________________
> > > Do you Yahoo!?
> > > Yahoo! Mail Plus - Powerful. Affordable. Sign up
> > now.
> > > http://mailplus.yahoo.com
> > >
> > > -------
> > > The MPLS-OPS Mailing List
> > > Subscribe/Unsubscribe:
> > http://www.mplsrc.com/mplsops.shtml
> > > Archive:
> > http://www.mplsrc.com/mpls-ops_archive.shtml
> >
> > I'm not sure if that would be an issue. First of all
> > you can bundles
> > some or all FEC
> > in one packet so it doesn't consume lot of BW.
> > Second, those messages
> > usually
> > only occured once unless you've lot of
> > topology/routing changes.
> >
> >
> > Carlos
> >
> >
> >
> >
>
> __________________________________________________
> Do you Yahoo!?
> Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
> http://mailplus.yahoo.com


-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml