[olug] dynamic routing for small business

neal rauhauser 402-301-9555 neal at lists.rauhauser.net
Sat Jul 26 19:08:12 UTC 2003


  There is no point to setting up dynamic routing if you're a leaf node
for an ISP. I have several customers with dyanmic routing - I'll
describe briefly.

  One customer has two Cisco 2620s at each location - a T1 ties two of
the 2620s together and the other pair have a backup VPN link between
them. EIGRP was the perfect thing for this network with half a dozen
subnets.

  One customer has two locations tied together via a single T1 and DSL
at each location. EIGRP here also - if the DSL drops on one end the
default route originated from the other location lets internet access
continue without the admin's intervention, and there is a VPN between
the two routers in case the T1 fails.

  One customer has a location with a wired and wireless path to its
upstream internet provider. The upstream used to be run by this crazy
brave Cisco admin who had customer routers participating in their OSPF
cloud to allow multihoming for customers without justifying an entire
/24 and using BGP. The multihomed location serves three remote sites
with dual wireless/wired service and we layered EIGRP over the VPN links
for dynamic support of their private address space.


  RIP is at times useful - broadly supported, not nearly as demonic as
its reputation when used in smallish networks without redundant links.
I've seen RIP v2 carry *eight hundred* routes in a topology with few
redudant links and actually behave itself most of the time.



 IGRP I've never seen except in lab environments, same goes for IS-IS.


 If you're starting from scratch just skip RIP and go for the hard stuff
- OSPF if its multivendor, EIGRP if you're impatient and its pure Cisco.




  




Brian Roberson wrote:
> 
> My short take on the thread: arcane != archaic
> 
> http://www.webster.com/cgi-bin/dictionary?va=archaic
> http://www.webster.com/cgi-bin/dictionary?va=arcane
> 
> dynamic routing is not archaic, it is arcane however.... if I
> interpreted that word correctly ;-)
> 
> my 2 cents on DRP's in small networks: it is simply overkill and not
> needed.
> 
> On Fri, 2003-07-25 at 15:37, Nick Walter wrote:
> > I'm going to respectfully disagree on the complexity of dynamic
> > routing.  Routing (static and dynamic) is an arcane topic from the
> > viewpoint of a small business.  While I'm sure very many of the people
> > on this list consider it an approachable and interesting topic, the
> > average small business doesn't have that level of staff talent on hand.
> > I've dealt with lots of admins, even ones at large companies, that get
> > confused dealing with static routing situations.  While it's possible a
> > mediocre admin could follow a HOWTO guide and get zebra up and running,
> > they would be left with a setup they didn't 100% understand.  And that
> > spells disaster the first time the setup fails and needs quick
> > troubleshooting.  I'm not trying to discourage anyone from learning
> > about this stuff, I'm just saying from a practical standpoint it should
> > be avoided unless one really knows what one is doing.  Otherwise it just
> > adds a complexity factor that makes extended network outages more
> > likely.
> >
> > I do stand corrected on my "spendy hardware" reference.  I've never
> > bothered to try it on a Linux-based router, I always leave those kind of
> > tasks to the  Cisco equipment.  And that stuff is spendy ;-)
> >
> > Nick Walter
> >
> >
> > On Fri, 2003-07-25 at 14:21, Matthew G. Marsh wrote:
> > > On Fri, 25 Jul 2003, Eric Penne wrote:
> > >
> > > Well - I was going to stay out of this one ... buuuut...
> > >
> > > > If it is so arcane then what is it's replacement?
> > >
> > > It is not arcane nor does it require $$s.
> > >
> > > http://www.policyrouting.org/PolicyRoutingBook/ONLINE/TOC.html
> > >
> > > and check out Chapter 7. There is some information on getting simple
> > > RIP/OSPF going. Also see both the Zebra and Bird projects along with the
> > > venerable gated.
> > >
> > > There are several scenarios where dynroutes are useful. Most notably when
> > > running a real VPN (consider that 2.6 wil have a fully policy routing
> > > compliant IPSec implementation). Also when monitoring link status. Many
> > > routing daemons notice link flaps very quickly and can send SNMPv3 traps.
> > >
> > > > > A small business with only one link to the internet does not need to
> > > > > muck around with dynamic routing.  This is a good thing, because dynamic
> > > > > routing is arcane, requires spendy hardware, and is easy to screw up.
> > > > >
> > > > > The only time a business needs to start worrying about dynamic routing
> > > > > is when they have multiple links to the internet.  In that case, dynamic
> > > > > routing becomes a necessity to have traffic re-routed around if one link
> > > > > fails.
> > > > >
> > > > > Nick Walter
> > > > >
> > > > > On Fri, 2003-07-25 at 11:33, Eric Penne wrote:
> > > > >> I was curious about what dynamic routing was.  I did a little googling
> > > > >> and found out what some of the acronyms I had seen in the past couple
> > > > >> of years meant.
> > > > >>
> > > > >> What I never did find is when is it useful.  I know it is useful for
> > > > >> large companies with lots of sites and network connections and stuff
> > > > >> but what about small single connection places?  Is it useful to build
> > > > >> a router with dynamic routing if you just have a single t1?  Does
> > > > >> dynamic routing help you and others to determine the best path for
> > > > >> packets?  Will running a router with dynamic routing help anybody else
> > > > >> if you don't actively participate?  These are quite being worded the
> > > > >> way I want them to.
> > > > >>
> > > > >> Ok here it goes again.  I assume to use dynamic routing effectively
> > > > >> you have to collaborate with other routers to find the best path.  If
> > > > >> I were to just set up a router on my t1 without actively collaborating
> > > > >> with my isp or next step up the chain, would it help anything?
> > > > >>
> > > > >> Eric
> > > > >>
>


More information about the OLUG mailing list