OSPF ECMP over long distance gig links

Unanswered Question
Nov 5th, 2009
User Badges:

Hi,


We have a MPLS network between to datacenter locations. Between the core nodes we have 1 Gbps Ethernet links transported via a DWDM network.


We are now in the process of upgrading the capacity between the datacenters.


The final solution is to implement 10 Gbps, but as a fast solution to have more capacity we are planning to install 1 more GigE link between to of the core nodes. This is a temporary solution to quickly implement more bandwidth.


See attached simplified network drawing.


We are planning to use OSPF ECMP on the links between core1 and core3. All nodes are 6500 with 67xx interfaces. The two links between core1 and core3 have different paths in the DWDM transport network, which will give a slight difference in latency.



My questions are:

1. Is this a safe configuration?

2. Is "per-destination load-sharing" default behavior on IOS based 6500 (12.33.SXI1)

3. Are there any issues we have to take into consideration before implementing?



Attachment: 
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Reza Sharifi Thu, 11/05/2009 - 09:29
User Badges:
  • Super Bronze, 10000 points or more
  • Cisco Designated VIP,

    2017 LAN

Hi Jorgen,


Since you just need more bandwidth, you can create an EtherChannel and bundle the 2 1Gig links together. This way you do not have to worry about another OSPF peering, additional IP subnet and also the interface hardware is handling the load balancing.


HTH

Reza

jorgensor Thu, 11/05/2009 - 14:52
User Badges:

An Etherchannel was an option to this solution. Though we want to have a pure L3 solution when running traffic over long distance.

iyde Fri, 11/06/2009 - 10:29
User Badges:
  • Silver, 250 points or more

Hi.


Then make a L3 etherchannel where you have no switchport command on the physical interfaces you will have i your channel and then put the IP address on the Pox interface.


HTH, Ingolf

Giuseppe Larosa Tue, 11/10/2009 - 05:12
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Jorgen,

using OSPF over the two links to have load balancing will mean using CEF based load balancing.

And it is good enough for normal traffic patterns.


CEF load balancing is based on:

IP SA exor IP DA exor seed hash


with an etherchannel IPv4 traffic is load balanced using:

IP SA exor IP DA.


I would stay with two parallel IP links.


I don't see why you should look for troubles trying to move to an etherchannel.


Edit:


>>My questions are:

1. Is this a safe configuration?


<< yes

2. Is "per-destination load-sharing" default behavior on IOS based 6500 (12.33.SXI1)


<< yes see above explanation of exor

3. Are there any issues we have to take into consideration before implementing?


<< No


Hope to help

Giuseppe


Actions

This Discussion