>Howard,
>You might be able to clarify the OSPF RFC for me on this one. RFC 2328
>talks about a routing table, and routing table entries (section 11). So
>does Doyle. My interpretation of this was that they are both referring to
>a routing table internal to OSPF, separate from the overall routing table
>that includes all statics, OSPF, etc etc routes.
>So if OSPF *doesn't* have it's own routing table as such... are they really
>both referring to the overall routing table? Or are they referring to a
>temporary structure that is set up and thrown out each time the SPF
>algorithm runs?
>
>JMcL
I haven't seen the IOS code, but of the OSPF implementations I know
about, there is only a temporary structure of intra-area routes.
Inter-area and external routes are persistent only as LSDB entries.
Just as a minor point, OSPF's routing table is more than
Dijkstra/SPF. The Dijkstra only handles intra-area routes. Since
intra-area routes are always preferred to inter-area and external, if
there is an intra-area route to a destination, the less preferred
type is excluded.
So what, you might very easily ask, does the second stage use to look
up intra-area routes? In the real world, there is a temporary
"routing table." Also in the real world, this would be too large to
keep around, given that it would largely be duplicated in the "main"
routing table (i.e., Routing Information Base). Yes, I can see there
would be interest in having the information for debugging, but my
feeling would be that most implementers think it's too costly.
From the Cisco point of view, if you want to play "what if I remove
this static" games, Netsys is intended for just that purpose. Not
cheap, I recognize.
The description of SPF in RFC2328 is conceptual, and the performance
of an OSPF implementation depends on how efficiently the real
computation is written. Dijkstra's algorithm is 40 years old, and
more efficient approaches exist. See, for example, Tony Lauck's
modification described in Perlman's _Interconnections_ (1st Ed), page
225.
A presentation at the last NANOG reviews current approaches to
millisecond-range IGP convergence:
http://www.nanog.org/mtg-0010/igp.html While the research here
involved IS-IS, it's still applicable. There is enough interest in
millisecond convergence that if there were spare resources in the
router, I'm afraid supporting this is a higher priority than
what-if's.
>
>---------------------- Forwarded by Jenny Mcleod/NSO/CSDA on 28/11/2000
>08:59 am ---------------------------
>
>
>"Howard C. Berkowitz" <[EMAIL PROTECTED]>@groupstudy.com on 28/11/2000 01:37:13
>am
>
>Please respond to "Howard C. Berkowitz" <[EMAIL PROTECTED]>
>
>Sent by: [EMAIL PROTECTED]
>
>
>
>To: [EMAIL PROTECTED]
>cc:
>
>
>Subject: Re: IP routing tables
>
>
>>Kevin,
>>Yes, the candidate routes sound like what I'm after. I expect you're
>right
>>about needing a separate command for each routing protocol. I have a
>>feeling that I may be asking for something that doesn't exist, at least
>not
>>for all protocols.
>>
>>I've already dug around the 'show ip ospf database' commands, and I can't
>>find one that actually shows the internal ospf routing table, but heaven
>>knows there are enough combinations that I could have missed plenty.
>'show
>>ip ospf database' by itself just shows a list of LSAs, which while useful
>>is not what I'm after in this case.
>>'show ip ospf border-routers' shows the router entries in the route table,
>>which is a start - but is there a way of seeing the network entries?
>>
>>And another nasty one - how about static routes (short of 'show run', of
>>course)?
>>
>>JMcL
>
>
>As far as I know, the only routing protocol that really keeps its own
>routing table is BGP.
>
>You might see OSPF's routes with a debug, but essentially, it
>computes routes from the LSDB, sends them to the routing table
>manager, and then forgets the specific intra-area routes. Inter-area
>and external routes do have LSDB entries that correspond to the
>route, but there isn't a routing table in the sense you describe.
>
>[snipped]
>>
>
>
>_________________________________
>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]