Guest blog by Frederic Trate, SP Product and Solutions Marketing
For those, who read my previous blog, you surely want to know more about Application Engineered Routing. As Segment Routing sits at the very heart of Application Engineered Routing, let me start explaining how the former works.
I’m going to draw an analogy between how IP networks route data traffic and how the airline companies transport luggage. The point here is not to get into technical details; for those interested in the nitty-gritty of Segment Routing, I recommend reading the following white paper:
Let’s take the example of a piece of luggage, which is traveling from Los Angeles to Berlin. There are multiple routes between Los Angeles and Berlin but let’s consider only 2 of them for the sake of simplicity:
- Route 1: Los Angeles – New York – Berlin
- Route 2: Los Angeles – Mexico – Madrid – Berlin
First scenario: the luggage is routed over the route with the fewest stops.
In our case, this is Route#1. A tag, with the final destination of Berlin, is attached to the luggage, which is then boarded on a flight to New York. Once in New York, the luggage is screened to determine the next flight it should be boarded on.
In this scenario, every piece of luggage is transported from Los Angeles to Berlin through New York so transport capacity can be an issue.
This is exactly how Layer 3 routing works today – IP traffic is routed over the shortest path between two nodes.
Second scenario: the luggage is routed over a specific route.
Let’s say it is routed over Route#2.
As this route is not the shortest one (e.g. it has more stops than Route 1), a specific luggage tag is created to identify Route#2. Since in the MPLS world a tag (MPLS label) has local significance only, each intermediate hop needs to be programmed to recognize the incoming tag and to know which next hop destination is corresponding and attach the outgoing tag to be processed by the next hop.
In Los Angeles, the (end to end) tag is attached and the luggage sent to Mexico. Once in Mexico, the tag is analyzed, next hop is identified to be Madrid and an outgoing tag attached. The outgoing tag uniquely identifies the end to end path and has a local significance for Madrid. In Madrid, the tag is analyzed and Berlin identified as destination and the outgoing tag attached to the bag.
As pre-provisioning and manual intervention at each hop (luggage tag swap) are required, you easily understand why this solution is not well suited for transporting hundred of thousands pieces of luggage over million possible end to end paths.
This is how MPLS Traffic Engineering works.
So what’s different with Segment Routing?
Let’s go back to our luggage analogy and see how a piece of luggage would be routed over the specific Route 2.
In Los Angeles, a global tag is attached to the luggage – it contains all the hops the luggage will go through up to its final destination. In our case, this means the global tag is printed with Mexico-Madrid-Berlin. Once in Mexico, the luggage is automatically routed to its next destination by simply looking at the next hop after Mexico, Madrid. Same process happens in Madrid.
Sounds familiar to you? Right, this is exactly how airline companies transport your luggage!
Good news is this is exactly how Segment Routing works. As segments are global and the path decided and encoded on top of the packet at ingress no specific programming needs to be done at intermediate hops. It is simple and scalable.
At this stage, you better understand how Segment Routing works and why it is so powerful to steer traffic over specific network paths.
Next step is to add a SDN controller. What about WAN Automation Engine (WAE)?
The new release of WAE, available in July 2015, supports Segment Routing and enables you to deliver new Use Cases, such as Low Latency and Avoidance, I’ve touched upon in my previous blog.
Listen in to Mike Thompson, General Manager WAN Orchestration, to get further insight:
Stay tuned … more to come on Application Engineered Routing solution.