- A virtual community of network engineers
 Home  BookStore  StudyNotes  Links  Archives  StudyRooms  HelpWanted  Discounts  Login
Re: (correction) Method and Process Scenario 5: OSPF [7:42139] posted 04/22/2002
[Chronological Index] [Thread Index] [Top] [Date Prev][Date Next] [Thread Prev][Thread Next]

>question embedded within:
>""Howard C. Berkowitz""  wrote in message
>>  >Howard,
>  > >        I think I see where you're going.  The default on the "DIO"
>>  >command is
>>  >applying an "E2" to the default as it is sent into the enterprise.  It
>>  >also
>>  >known that by order of preference that "E2" routes are least preffered.
>>  >So based
>>  >on your hint.. I'm thinking making use of the "metric-type" parameter to
>>  >make
>>  >the default-route an "E1" metric which would provide known route info
>>  >into the
>>  >ISP's network.  Nope..this isn't it.
>>  STOP!  Using E1 is the answer, although I don't think you have the
>>  reason quite right.  On the default-information originate command,
>>  use metric-type 1 and an equal metric on both routers.
>  > E1 considers the combined internal and external metric.  If you make
>>  the external metrics equal, traffic in your network will go to the
>>  closest exit.  If the network topology is reasonably well designed
>>  with the placement of your gateways, this should give approximate
>>  sharing of both internal resources and the ISP links.
>I'm wondering how many readers of this thread fooled themselves by thinking
>that the idea was to ensure per packet load sharing out the two ISP links?
>which no doubt leads to suboptimal routing for a significant portion of
>traffic, if my mental picture is correct.
>aren't the two goals - equal load sharing and optimal routing - mutually
>exclusive here?

In practice, no, if you think carefully where you place the ISP 
gateways.  Typically, they should be at opposite geographical ends of 
your network, near heavy concentrations of users.  That often causes 
load sharing by pure use distribution.  It's certainly not per-packet 
between multiple routers.  It's more per-destination for individual 
routers, CEF of course giving even better results than fast switching.

The optimal routing to which we are referring is internal, not 
external. It presupposes the ISP links are of equal capacity.

Message Posted at:
FAQ, list archives, and subscription info:
Report misconduct and Nondisclosure violations to abuse@xxxxxxxxxxxxxx