From: James (james@towardex.com)
Date: Thu Jul 15 2004 - 16:09:18 GMT-3
On Thu, Jul 15, 2004 at 02:37:41PM -0400, Howard C. Berkowitz wrote:
> At 9:58 AM -0700 7/15/04, Joseph D. Phillips wrote:
> >There are many political or economic reasons why optimal routing isn't
> >always possible or desired.
> >
> >The job of a network engineer, as I see it, is to ensure data flows
> >according to the requirements and directives of the organization. You
> >may have great ideas about how to improve the flow of traffic, but you
> >can slam into bureaucratic walls trying to implement your designs.
> >
> >Can I get a witness? Scott? Brians? Howard?
>
> Thanks -- I've been meaning to respond.
>
> Optimal routing has no single meaning. For example, there is the
> concept variously called hot potato or closest exit routing. Examples
> here would be an ISP handing off the packet to the first exit it can
> find, so the workload imposed on the ISP backbone is minimized.
>
> In like manner, in an OSPF totally stubby area with multiple ABRs,
> you only have information to get the packet to the closest ABR -- you
> are only locally optimizing within the area. Broadening the scope of
> OSPF to ISPs, you can use Type 1 externals to load share within your
> domain, but, without traffic engineering, you have no control of
> optimization once the packet leaves your domain.
>
> It's been my experience that people tend to overemphasize routing by
> metric, etc., in that there usually aren't so many different possible
> routes that the best metric really makes much different. Topology,
> particularly hierarchical topology, tends to be far more important.
Yep. Speaking of hot potato, then there comes the cold potato with
MED (and may be few other attributes, depending on how ISP wants)
tweaks on routes. Unfortunately, for egress traffic leaving your
network, cold potato through the industry has proven not a lot useful
than overemphasized routing and more complexity. While cold potato
on ingress traffic, if you are a transit customer of your adjacent
AS, is a good idea, I'd rather spend more money on upgrading peering
and transit capacity than on my own transports instead -- which is
means hot potato'ing the egress traffic.
There really is no meaning in optimal routing I agree. Whether you
use hot potato or cold potato, the difference is minimal in most
cases. In fact, given that majority of the internet structure is
based upon "best effort delivery" belief, I would rather let my
peer decide how it wants to deliver that traffic to its own customer
edge as early as possible, than playing metrics over my transport
circuits.
-J
-- James Jun TowardEX Technologies, Inc. Technical Lead Network Design, Consulting, IT Outsourcing james@towardex.com Boston-based Colocation & Bandwidth Services cell: 1(978)-394-2867 web: http://www.towardex.com , noc: www.twdx.net
This archive was generated by hypermail 2.1.4 : Sun Aug 01 2004 - 10:11:56 GMT-3