Whoops! When I posted Excerpt 1, I missed the part in the first sentence that
seems to imply that frame-relay interfaces and multipoint interfaces disable
split-horizon BY DEFAULT (when using TCP/IP).
When I took ICRC so many years ago, I remember discussing split-horizon but
don't recall any explicit mention that split-horizon is disabled by default
for the IP routing protocols (RIP and IGRP). I remember discussing this issue
as well in CIT but I don't think the exception to the rule was noted
either--but because I wasn't aware of it, it may have been something that went
in one ear and out the other.
It's interesting to note that in the current ICND curriculum, I know the
section that talks about configuring subinterfaces indicates that "multipoint
subinterfaces act as NBMA network so they do not resolve the split horizon
issue". I don't have the entire official ICND curriculum available, but I
don't notice anything specifically mentioning the exception.
So, in summary - we're both right with respect to multipoint interfaces:
With multipoint (or physical) interfaces: split horizon is ENABLED by default
unless you are routing IP across the link. Then it is DISABLED. If you're
routing using RIP, IGRP, or EIGRP to route IP, it's DISABLED. If you're
routing IPX using RIP, EIGRP, NLSP, it's ENABLED. It's also enabled if
routing AppleTalk.
With point-to-point subinterfaces, split horizon is ENABLED but isn't much of
an issue since it's a point-to-point link.
-- Leigh Anne
> -----Original Message-----
> From: Brian Lodwick [mailto:[EMAIL PROTECTED]]
> Sent: February 27, 2001 11:26 AM
> To: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> Subject: RE: FR + Poison Reverse...
>
>
> Leigh Anne Chisholm writes:
> Brian L:
>
> I think you've got it backwards as to when split-horizon operates on an
> interface. By default, all Cisco serial interfaces are multipoint unless
> specifically configured to be point-to-point. With multipoint, you'd want
> split-horizon enabled if you were using a broadcasting routing protocol.
>
> --Excerpt 1--
>
> "Note: For TCP/IP, Cisco routers can disable split-horizon limitations on
> all frame relay interfaces and multipoint subinterfaces and do this by
> default. However, split-horizon cannot be disabled for other protocols like
> IPX and AppleTalk. These other protocols must use subinterfaces if dynamic
> routing is desired."
>
>
> Brian's reply:
> You have just cut from a cisco document that proves I am accurate. This says
> all interfaces are by default multipoint unless specifically configured
> point-to-point. This is talking about physical interface configuration not
> multipoint subinterface configuration, and you have just proved I am correct
> that Physical interfaces configured for frame-relay by default disable
> split-horizon. Also if you want a distance-vector protocol to work correctly
> on a Frame-relay hub & spoke model split-horizon must be disabled on the
> hub(unless the hub is configured for point-to-point to each spoke)
>
> >>>Brian
>
> >From: "Leigh Anne Chisholm" <[EMAIL PROTECTED]>
> >Reply-To: <[EMAIL PROTECTED]>
> >To: "Brian Lodwick" <[EMAIL PROTECTED]>, <[EMAIL PROTECTED]>,
> ><[EMAIL PROTECTED]>, <[EMAIL PROTECTED]>
> >Subject: RE: FR + Poison Reverse...
> >Date: Tue, 27 Feb 2001 11:03:21 -0700
> >
> >If you don't want to read all of my quotes, look for the portion contained
> >therein enclosed by double asterisks (**).
> >
> >From Cisco's site:
> >
> >"Normally, routers that are connected to broadcast-type IP networks and
> >that use distance-vector routing protocols employ the split
> >horizon mechanism to reduce the possibility of routing loops. Split horizon
> >blocks information about routes from being advertised by
> >a router out any interface from which that information originated. This
> >behavior usually optimizes communications among multiple
> >routers, particularly when links are broken. However, with nonbroadcast
> >networks, such as Frame Relay and SMDS, situations can arise
> >for which this behavior is less than ideal. For these situations, you might
> >want to disable split horizon. **This applies to IGRP
> >and RIP.**" This excerpt can be found at:
> >
>
>http://www.cisco.com/univercd/cc/td/doc/product/software/ios11/cbook/ciproute
.htm#xtocid16743169
> >
> >A few weeks ago, this same discussion was brought up. Someone posted the
> >following link outlining "Hub And Spoke Frame Relay Sample
> >Configuration - Dynamic (OSPF) IP Routing". Here's an excerpt from the
> >link:
> >
> >"In general, it is good practice to use subinterfaces for partially-meshed
> >frame relay networks. A frame relay network designed with
> >subinterfaces scales much easier to future expansion. Referring to the
> >example, subinterfaces allow routing updates to exchange
> >between Boston and Chicago through Atlanta. **Without subinterfaces, Boston
> >is unable to receive routing updates from Chicago and
> >vice versa creating a condition known as split-horizon.**" This excerpt
> >can be found at:
> >
>
>http://www.cisco.com/warp/public/779/smbiz/service/configs/framerelay/fr_ip_o
spf.htm
> >
> >Now this is in contrast to what I know of OSPF and in contrast to the first
> >quote I've included. Quite frankly, I would think that
> >the problems encountered in the Frame Relay example have more to do with
> >the non-broadcast nature of Frame Relay than split-horizon
> >so I think whomever wrote this may have misunderstood why subinterfaces
> >enable routing updates--and I think perhaps this is why
> >there's so much confusion over whether or not split-horizon plays any role
> >in OSPF.
> >
> >Brian L:
> >
> >I think you've got it backwards as to when split-horizon operates on an
> >interface. By default, all Cisco serial interfaces are
> >multipoint unless specifically configured to be point-to-point. With
> >multipoint, you'd want split-horizon enabled if you were using
> >a broadcasting routing protocol. Here's two excerpts from Cisco's site:
> >
> >--Excerpt 1--
> >
> >"Note: For TCP/IP, Cisco routers can disable split-horizon limitations on
> >all frame relay interfaces and multipoint subinterfaces
> >and do this by default. However, split-horizon cannot be disabled for other
> >protocols like IPX and AppleTalk. These other protocols
> >must use subinterfaces if dynamic routing is desired."
> >
> >--Excerpt 2--
> >
> >"Cisco serial interfaces are multipoint interfaces by default unless
> >specified as a point-to-point subinterface. Though less common
> >than point-to-point subinterfaces, it is possible to divide the interface
> >into separate virtual multipoint subinterfaces."
> >
> >"Multipoint interfaces/subinterfaces are still subject to the split-horizon
> >limitations as discussed above. All nodes attached to a
> >multipoint subinterface belong to the same network number. Typically,
> >multipoint subinterfaces are used in conjunction with
> >point-to-point interfaces in cases where an existing multipoint frame relay
> >cloud is migrating to a subinterfaced point-to-point
> >network design. A multipoint subinterface is used to keep remote sites on a
> >single network number while slowly migrating remote
> >sites to their own point-to-point subinterface network."
> >
> >"Figure 4 shows serial 0.1 as a multipoint subinterface connecting to three
> >different locations. All devices on the multipoint
> >subinterface belong to the same network number (100). Site E has migrated
> >off of the multipoint network to its own point-to-point
> >subinterface network (200). Eventually, all remote sites can be moved to
> >their own point-to-point subinterface networks and the
> >multipoint subinterface will not be necessary."
> >
> >Both of these excerpts can be found at:
> >
> >http://www.cisco.com/warp/public/779/smbiz/service/knowledge/wan/subifs.htm
> >
> >
> > -- Leigh Anne
> >
> > > -----Original Message-----
> > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
> > > Brian Lodwick
> > > Sent: February 27, 2001 10:21 AM
> > > To: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> > > Subject: RE: FR + Poison Reverse...
> > >
> > >
> > > OSPF uses split-horizon? I don't think this is correct. I hate to be
> >quick
> > > to reply but there is no reason for OSPF to use split horizon. OSPF
> >sends
> > > updates to all adjacent neighbors, and in NBMA , and Broadcast network
> >types
> > > the DR forwards the updates on.
> > > Split-horizon is used to keep routing loops from happening for a
> > > Distance-vector protocol not a link state protocol.
> > > There is an issue split-horizon presents when using a distance-vector
> > > protocol in a frame-relay hub spoke topology. As you can imagine (if you
> > > know how a distance-vector protocol diseminates updates) the hub will
> >need
> > > to send the update back out of the interface it received it on so that
> >the
> > > other spokes will receive the update. This will not be allowed to happen
> >if
> > > split-horizon is enabled. The spokes will not be an issue.
> > >
> > > By default:
> > > Physical interface frame-relay setup split-horizon will be disabled.
> > > Subinterface frame-relay setup split-horizon will be enabled.
> > >
> > > Also keep in mind you cannot disable split-horizon on IPX RIP.
> > >
> > > Summation when you have a hub spoke frame-relay topology and you are
> >using a
> > > distance-vector routing protocol you need to disable split-horizon on
> >the
> > > hub. (which does leave you secceptible to the issue split-horizon was
> > > designed to fix)
> > >
> > > Or use a link-state protocol.
> > >
> > > Or you could setup point-to-point subinterfaces on the hub for each
> >spoke.
> > >
> > > >>>Brian
> > >
> > >
> > > >From: "Maness, Drew" <[EMAIL PROTECTED]>
> > > >Reply-To: "Maness, Drew" <[EMAIL PROTECTED]>
> > > >To: "'Z'" <[EMAIL PROTECTED]>, Brian <[EMAIL PROTECTED]>,
> > > >[EMAIL PROTECTED]
> > > >Subject: RE: FR + Poison Reverse...
> > > >Date: Tue, 27 Feb 2001 08:36:57 -0800
> > > >
> > > >OSPF does use Split horizon. I don't think OSPF uses PR because PR
> >sets
> > > >the
> > > >route to infinity and I'm not sure what an ' infinity' cost would mean
> >in
> > > >OSPF.
> > > >
> > > >-----Original Message-----
> > > >From: Z [mailto:[EMAIL PROTECTED]]
> > > >Sent: Tuesday, February 27, 2001 12:47 AM
> > > >To: Brian; [EMAIL PROTECTED]
> > > >Subject: Re: FR + Poison Reverse...
> > > >
> > > >
> > > >But I thought that I'd seen or heard of using split horizon with
> > > >OSPF...maybe I'm thinking of Frame Relay...long day I guess...
> > > >
> > > >************************************************************
> > > >This has been an Eyez Only streaming e-mail broadcast...We are
> >watching.
> > > >
> > > >NetEyez ~ CCNP, CCDA
> > > >
> > > >----- Original Message -----
> > > >From: "Brian" <[EMAIL PROTECTED]>
> > > >To: "Z" <[EMAIL PROTECTED]>
> > > >Cc: <[EMAIL PROTECTED]>
> > > >Sent: Tuesday, February 27, 2001 1:16 AM
> > > >Subject: Re: FR + Poison Reverse...
> > > >
> > > >
> > > > >
> > > > >
> > > > > split horizon and PR are both associated with distance vector
> >protocols,
> > > > > OSPF is a link state protocol.
> > > > >
> > > > > Brian
> > > > > On Mon, 26 Feb 2001, Z wrote:
> > > > >
> > > > > > Hey Group,
> > > > > > I recently pondered something while at work and nobody could
> >give
> > > >me
> > > >a
> > > > > > strait answer. Actually I feel a little embarrassed asking this
> >due to
> > > >me
> > > > > > being an NP and I feel like I should know this. Guess this type of
> > > >thing
> > > > > > doesn't come up much around me. Question is: Can poison reverse
> >(PR)
> > > >be
> > > >used
> > > > > > with OSPF? I know split horizon is used with it but I just cant
> >see
> > > >why/how
> > > > > > poison reverse would. My definition of PR is that it sets the link
> >to
> > > >the
> > > > > > max hop count and deems it unreachable, hence the term poison. I
> >can't
> > > >see
> > > > > > how this would work with OSPF because it doesn't use a hop count.
> > > >Maybe
> > > >I'm
> > > > > > confused about PR. Does it set the link to the highest metric, and
> >not
> > > >hop
> > > > > > count? Maybe I'm just used to hearing about PR in discussions of
> >RIP
> > > >that
> > > > > > I'm assuming it set the hop count to the highest and has nothing
> >to do
> > > >with
> > > > > > metric. Any clarity would help, thanks all...
> > > > > >
> > > > > > ...sorry for the rambling...
> > > > > >
> > > > > > ************************************************************
> > > > > > This has been an Eyez Only streaming e-mail broadcast...We are
> > > >watching.
> > > > > >
> > > > > > NetEyez ~ CCNP, CCDA
> > > > > >
> > > > > >
> > > > > >
> > > > > > _________________________________
> > > > > > FAQ, list archives, and subscription info:
> > > >http://www.groupstudy.com/list/cisco.html
> > > > > > Report misconduct and Nondisclosure violations to
> >[EMAIL PROTECTED]
> > > > > >
> > > > >
> > > > >
> > > > > -----------------------------------------------
> > > > > I'm buying / selling used CISCO gear!!
> > > > > email me for a quote
> > > > >
> > > > > Brian Feeny,CCDP,CCNP+VAS Scarlett Parria
> > > > > [EMAIL PROTECTED] [EMAIL PROTECTED]
> > > > > 318-222-2638 x 109 318-222-2638 x 101
> > > > >
> > > > > Netjam, LLC http://www.netjam.net
> > > > > 1401 Oden St.
> > > > > Suite 18
> > > > > Shreveport, LA 71104
> > > > > Fax 318-221-6612
> > > > >
> > > > >
> > > >
> > > >_________________________________
> > > >FAQ, list archives, and subscription info:
> > > >http://www.groupstudy.com/list/cisco.html
> > > >Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> > > >
> > > >_________________________________
> > > >FAQ, list archives, and subscription info:
> > > >http://www.groupstudy.com/list/cisco.html
> > > >Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> > >
> > > _________________________________________________________________
> > > Get your FREE download of MSN Explorer at http://explorer.msn.com
> > >
> > > _________________________________
> > > FAQ, list archives, and subscription info:
> >http://www.groupstudy.com/list/cisco.html
> > > Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> >
>
> _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
