Bill,
I can give you two different answers. The first one is like backing up, the  
second one is not afraid of your "punch".
 
1) should politics follow suit ! And they certainly will. Example: When the  
telco's became private companies
Least Cost Routing was the big topic. Multiple distance zones were defined  
as to affect billing. But soon companies realized that they were unable to  
implement such policy and that they might get in big trouble if the per  call 
detailed bill weren't correct. So all of a sudden, the flat rate was  born and 
telco's who eventually were capable of doing distance-zone-conform  charging ( 
I 
doubt that there were some) concurred i.e. adopted the flat rate  too. Will 
say: policy will adopt to what is feasible. 
 
2) I would run 99 times my going-beyond-Dijkstra algorithms (anyway) as to  
validate all 300 links and each of them wrt to both directions how well they 
can  be used. Some A or some B who want's to discriminate using the link A<-->  
B as to deny servicing 10.0.0.0/8 should advertise this policy, however  well 
scoped. When a packet arrives it will have to be screened for  10.0.0.0/8. 
When I said only 3 table offset lookups are needed then to retrieve  the 
complete 
ranking info regarding all adjacent links wrt to the particular  destination 
location from best next hop to imposible because it entered a dead  end. This 
ranking info may eventually include such discriminating data. (but  others as 
well: the actual ranking order may reflect the current time as to do  
time-of-day sensitive routing for instance !
 
Try yourself time-of-day sensitive routing !
 
Heiner
 
 
In einer eMail vom 23.12.2008 15:25:38 Westeuropäische Normalzeit schreibt  
[email protected]:

On Tue,  Dec 23, 2008 at 3:35 AM,  <[email protected]> wrote:
> In  einer eMail vom 23.12.2008 00:16:53 Westeuropäische Normalzeit  schreibt
> [email protected]:
>
> On Mon, Dec 22, 2008 at 5:55  PM,  <[email protected]> wrote:
>> I think there are  enough folks around who are both experts in BGP and
>>  OSPF.
>> (1):
>> My question is, what sort of  policy/QoS/etc. routing can only be enabled
>> by a distance vector  protocol

> Address and direction sensitive link state. That is, for  some
> combinations of source, destination and direction of transit the  link
> is up, but for other combinations of combinations of  source,
> destination and direction of transit the link is  down.
>
> And so can link.state protocols, right ? !
> They  just do it in a different way. Also, you may color the nodes and links
>  such that you can even get Multiple Topologies (see OSPF-MT).

Go try  it.

Build yourself a random network of 100 nodes with a total of 200  to
300 links between those nodes. For each direction on each link,  assign
10 or so ranges of source+destination address for which the link  is
considered "up." Let me give an example of what I mean by  that:

Network:   A-B

Link direction AB is UP for  source 10.0.0.0/8, destination any
Link direction BA is UP for source any,  destination 10.0.0.0/8
Links AB and BA are DOWN for all other combinations  of source and destination

Once you've built your 100-node network "on  paper," do the following:

1. Calculate the size of your link-state  RIB.

2. Formulate an algorithm, presumably using Dijkstra, that  processes
that RIB into a next-hop FIB at a given node.

3. Calculate  the size of the resulting next-hop FIB.


What you'll find when you  do this is that the addition of direction
and source address to the RIB may  square its size versus a comparable
distance-vector RIB. Worse, your  Dijkstra calculation won't be able to
process out the impact of the source  address on the FIB calculation so
you'll have to carry the source address  into the FIB as well. This may
square the size of the FIB you need to carry  versus a distance vector
protocol.

With DV, the source address  factor is subtractive. It's handled by
reducing the database entries  propagated over particular links,
removing the ones that aren't acceptable  for that link. With
link-state it's multiplicative, vastly increasing the  database size.

Regards,
Bill Herrin


-- 
William D.  Herrin ................ [email protected]  [email protected]
3005 Crane  Dr. ...................... Web: <http://bill.herrin.us/>
Falls  Church, VA 22042-3004


 
_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to