Loop Free Alternate: IP Fast Reroute (FRR)

We can use the IPv4 Loop-Free Alternate Fast Reroute feature to reduce the Outages duration when the primary uplinks fail on a router.

When a link or a router fails, routing algorithms like ISIS or OSPF compute new routes that take into account the failure. The time taken for computation is called routing transition. Until the transition is complete and all routers are converged on a common view of the network, the connectivity between the source and destination pairs is interrupted. Routing transition depends on the numbers of routers present in that area. The bigger the area; more the time is. This may impact on revenue of the customer also. We can use the IPv4 Loop-Free Alternate Fast Reroute feature to reduce the routing transition time to less than 50 milliseconds using a pre-calculated alternate next hop. When a router is notified of a link failure, the router immediately switches over to the repair path to reduce traffic loss.

The repair path computation is done by the Intermediate System-to-Intermediate System (IS-IS) or OSPF and the resulting repair paths are sent to the Routing Information Base (RIB). The repair path installation is Forwarding Information Base(FIB).

Eigrp has it's own Loop Free Alternate.
Click Here To Read Rest Of The Post...

Segment Routing Traffic Engineering

In my previous post of Segment Routing: Alternate of LDP and RSVP, already discussed about the advantages of using SR w.r.t LDP. The utmost advantage of using Segment Routing is to steer the traffic which is normally called “Traffic Engineering”.

In case of MPLS we called Traffic Engineering as MPLS-TE but in case of Segment Routing, it could be SR-TE (Segment Routing Traffic Engineering Tunnel. RSVP is required to build the MPLS-TE tunnels along with LDP but in Segment Routing, all you need is the Segment IDs only. It is free from heavy RSVP signaling.

There are different ways to create Segment Routing Traffic Engineering Tunnels:-
1. Manually Create SR-TE tunnel and explicitly define the path. This is equivalent to MPLS-TE tunnel and define explicitly path by giving the next hop addresses.

2. Manually Create SR-TE tunnel but for path computation use PCE (Path Control Element). In this case headend router will become the PCC (Path Computation Client) and some software which sits outside of the network but has full view of topology will become a PCC. PCEP (Path Computation Element Protocol) is used between PCC and PCE. For reference http://tools.ietf.org/html/draft-sivabalan-pce-segment-routing-00

3. Dynamically Create SR-TE tunnel and compute path by using PCE. PCE is having API where user can push the headend and tailend routers information with some kind of SLAs. PCE will dynamically create the tunnel push it to router (PCC). Once this is done, PCC will ask for dynamic path also. For reference http://www.enog.org/presentations/enog-6/201-SR_ENOG.pdf

Click Here To Read Rest Of The Post...

Segment Routing: Alternate of LDP and RSVP - Test Results

In continuation to my previous post "Segment Routing: Alternate of LDP and RSVP", I would like to bring some testing results which really help me to understand better the behavior of segment routing control plane without using LDP and RSVP. In the testing environment, every router is allocated one node segment per PE.


Test Scenario 1:- Segment Routing Packets are sent across the network using multiple ECMP
This is the simplest one and works as per your IGP metric.

How does it differ from standard MPLS?
The only difference is that segment routing doesn't require any LDP to build its control plane. In MPLS labels are generated for each and every prefix in IGP or only for loopbacks. In segment routing, per PE is having single label which is called node segment.



Test Scenario 2:- Segment Routing Packets are sent across the network using single IGP Path
If destination has two ECMP paths, packets are always sent to both the paths. But this can be changed by tweaking the IGP metric and forcefully we can utilize single path instead of two. By using segment routing we are imposing two labels of different segment nodes and lookup is done according to that. By doing this we are only using a single path even though we have two ECMP paths are available.

How does it differ from standard MPLS and RSVP-TE?
If the same has to be done in MPLS, we need to enable RSVP-TE to play around it. RSVP-TE with LDP requires lot of planning in terms of bandwidth reservation and adding more complexity in terms of operations. But in SR, you don't require anything to get this done.



Test Scenario 3:- Segment Routing Packets are sent across the network not using IGP Path
Labels has been imposed according to the path required to forward the packets.
How does it differ from standard MPLS and RSVP-TE?
To get this done, we require LDP and RSVP-TE. We need to create the TE tunnels. Explicitly define the tunnel path. By doing this we are only adding the complications for the operation team. In segment routing, the all we are adding only node segment IDs in the label stack.




Click Here To Read Rest Of The Post...