Policy Routing, raise OSPF Cost, or am I totally off base?
in [Cisco]
|
Prev: Is the crypto map order important ? Or is that between ezvpn and l2tp/ipsec ?
Next: MPLS & IPSEC VPN
From: trbodsl on 22 Apr 2008 15:41 Here's a small diagram of a portion of a network. http://dzcad90.home.mchsi.com/diagram.jpg The two routers depicted on the left side of the diagram are in the same rack in our DR datacenter. The router on the right is at our main datacenter. There are other connections going into both the 2611XM and the 7206VXR, but are not relevent to this question. Right now, I have OSPF routing enabled throughout the entire network, and all the links pictured are in the same area. All traffic from the main datacenter rides the 9MB MPLS pipe. The T1 from the main datacenter to the DR datacenter sits idle at this time and herein is where the problem lies. The DR datacenter is also one of our branches we conduct business in, and we have workstations on the 1.2.1.0/24 network. The servers in the DR Datacenter are on the 1.2.2.0/24 network. When the 9MB pipe is congested with SAN or IBM iSeries replication traffic, we can experience some slowdown on our workstations. I've been going over ways to get workstation traffic (1.2.1.0/24 <-> 1.1.1.0/24) to traverse the lone T1, and DR Replication traffic (1.2.2.0/24 <-> 1.1.1.0/24) to traverse the 9MB pipe. I thought of using policy routing on both the 2611 and the 7206, but then the 7206 would send all traffic sourced from 1.1.1.0 to the next hop, which would be specified as the 2611XM. I also tried increasing the OSPF cost on the F0/0 Ethernet Interface to 65534, thinking that by making this a higher cost, the T1 would then be perferred and this didn't work either. Am I on the right track or can this not be done?
From: Bod43 on 23 Apr 2008 07:45 On 22 Apr, 21:41, trbo...(a)gmail.com wrote: > Here's a small diagram of a portion of a network. > > http://dzcad90.home.mchsi.com/diagram.jpg > > The two routers depicted on the left side of the diagram are in the > same rack in our DR datacenter. The router on the right is at our main > datacenter. There are other connections going into both the 2611XM and > the 7206VXR, but are not relevent to this question. > > Right now, I have OSPF routing enabled throughout the entire network, > and all the links pictured are in the same area. All traffic from the > main datacenter rides the 9MB MPLS pipe. The T1 from the main > datacenter to the DR datacenter sits idle at this time and herein is > where the problem lies. > > The DR datacenter is also one of our branches we conduct business in, > and we have workstations on the 1.2.1.0/24 network. The servers in the > DR Datacenter are on the 1.2.2.0/24 network. When the 9MB pipe is > congested with SAN or IBM iSeries replication traffic, we can > experience some slowdown on our workstations. > > I've been going over ways to get workstation traffic (1.2.1.0/24 <-> > 1.1.1.0/24) to traverse the lone T1, and DR Replication traffic > (1.2.2.0/24 <-> 1.1.1.0/24) to traverse the 9MB pipe. > > I thought of using policy routing on both the 2611 and the 7206, but > then the 7206 would send all traffic sourced from 1.1.1.0 to the next > hop, which would be specified as the 2611XM. I also tried increasing > the OSPF cost on the F0/0 Ethernet Interface to 65534, thinking that > by making this a higher cost, the T1 would then be perferred and this > didn't work either. > > Am I on the right track or can this not be done? Policy based routing is suitable for your requirements. You can match traffic against an access-list and route it depending on whether it matches or not. The access list can be any IP Standard or Extended access-list and so you can match source, dest, port, whatever. You can arrange to have the routing fall back to using the routing tables in the event that PBR fails. i.e. next hop unreachable.
From: trbodsl on 23 Apr 2008 11:37
On Apr 23, 6:45 am, Bo...(a)hotmail.co.uk wrote: > On 22 Apr, 21:41, trbo...(a)gmail.com wrote: > > > > > > > Here's a small diagram of a portion of a network. > > >http://dzcad90.home.mchsi.com/diagram.jpg > > > The two routers depicted on the left side of the diagram are in the > > same rack in our DR datacenter. The router on the right is at our main > > datacenter. There are other connections going into both the 2611XM and > > the 7206VXR, but are not relevent to this question. > > > Right now, I have OSPF routing enabled throughout the entire network, > > and all the links pictured are in the same area. All traffic from the > > main datacenter rides the 9MB MPLS pipe. The T1 from the main > > datacenter to the DR datacenter sits idle at this time and herein is > > where the problem lies. > > > The DR datacenter is also one of our branches we conduct business in, > > and we have workstations on the 1.2.1.0/24 network. The servers in the > > DR Datacenter are on the 1.2.2.0/24 network. When the 9MB pipe is > > congested with SAN or IBM iSeries replication traffic, we can > > experience some slowdown on our workstations. > > > I've been going over ways to get workstation traffic (1.2.1.0/24 <-> > > 1.1.1.0/24) to traverse the lone T1, and DR Replication traffic > > (1.2.2.0/24 <-> 1.1.1.0/24) to traverse the 9MB pipe. > > > I thought of using policy routing on both the 2611 and the 7206, but > > then the 7206 would send all traffic sourced from 1.1.1.0 to the next > > hop, which would be specified as the 2611XM. I also tried increasing > > the OSPF cost on the F0/0 Ethernet Interface to 65534, thinking that > > by making this a higher cost, the T1 would then be perferred and this > > didn't work either. > > > Am I on the right track or can this not be done? > > Policy based routing is suitable for your requirements. > You can match traffic against an access-list and route > it depending on whether it matches or not. The access > list can be any IP Standard or Extended access-list > and so you can match source, dest, port, whatever. > You can arrange to have the routing fall back to > using the routing tables in the event that PBR fails. > i.e. next hop unreachable.- Hide quoted text - > > - Show quoted text - Wow.. I was having a major mental block. As soon as I tied this to an extended access list it works like a champ. Shutting down the T1 interface automatically falls back to the OSPF routing table. This is perfect. I was just thinking and making this too hard in my head. Thanks for the advice! |