Segment Routing Based MPLS Vs Classic MPLS

Previous post has already covered why segment routing is required with all it's basic information. This post is mainly focusing on difference between SR Based MPLS Networks Vs Classic MPLS Networks.

	Segment Routing Based MPLS	Classic MPLS
MPLS Transport	IGP	IGP + LDP
IGP/LDP synchronization	Not Required	Required and Added More Latecny in Convergence
50msec FRR	IGP	IGP + RSVP-TE
Extra TE states to support FRR	No extra state	Extra states to manage
Optimum backup path	Yes	No
ECMP-capability for TE	Inbuilt	No
TE state only at headend	Yes	No (n^2 problem at midpoint)
Seamless Interworking with classic MPLS	Yes	NA
SDN Support	Yes	No
Routing	Constraint Based (Source can Define)	Destination Based + RSVP-TE
Link information	(Bandwidth, IGP metric, TE metric, SRLG ) is flooded throughout the IGP domain	No
Path Calculation	CSPF or By Using Centralized Controller	IGP + RSVP-TE
Scalability	High	Low
Operations and Troubleshooting	Low	High

And here is the rest of it.

Layer 3 MPLS VPN with Segment Routing - Nodal Segment

Traditionally Layer 3 MPLSVPN service requires two labels; VPNv4 Label and LDP (Transport Label). LDP is used to distribute the transport labels within service provider domain. But let’s see how Layer 3 MPLSVPN service will work in case of Segment Routing.

Below is the topology which has configured IGP and MP-iBGP for Service Provider network. CE1 and CE2 are two sites of same customer which are part of same vrf. This service provider core is free from LDP.

Segment Routing is used for transport label however MP-iBGP is used for VPNv4 labels. On every SR capable router, Loopback interface is configured as Node segment ID which is also called as Prefix segment ID.

Label 200 is Node/Prefix SID of PE2 and is advertised in the IGP protocol as SR label which is part of SRGB. For PE1 reaching CE2; BGP next hop is a PE2 loopback. PE2 loopback is flooded as a label of 200 in the IS-IS sub-TLV extension or OSPF Opaque LSA.

PE1 pushes label 200 as transport label and 500 as inner label.
PE1 and P1 don’t change the transport label and send the packet to the P2. P2 receives an implicit null label for the loopback of PE2, P2 does PHP (Penultimate Hop Popping), and thus only the VPN label is sent to the PE2.

Basics Of Segment Routing

Segment Routing is a new technology that will increase the benefit to IP and MPLS networks. It is an alternate to LDP and RSVP which means without using the LDP and RSVP we still can generate and distribute the transport labels and steer the traffic without using the RSVP signalling. Segment routing is based on label switching but for labels generation and distribution LDP and RSVP is not used. Segment Routing is an extension to an IGP (OSPF/ISIS). Labels are called segments in Segment Routing.

As per IETF draft “Segment Routing (SR) leverages the source routing paradigm. A node steers a packet through an ordered list of instructions, called segments. A segment can represent any instruction, topological or service-based.

Segment routing uses IP control plane but MPLS and IPv6 as data plane for its operation. There are two main components of Segment Routing:-
1. Control Plane:- Generation and distribution of transport labels across the Segment Routing domain by using IGP (OSPF/ISIS)
2. Data Plane:- Add the labels (Segments) on the packet header

Different Type of Data Plane Operation supported by Segment Routing
1. Continue:- Forwarding action based on active segment
2. Push:- Add a segment to the SR header of the packet and set that as Active Segment
3. Next:- Mark the next segment as the active segment and execute the instruction encoded by the new active segment

Comparison between MPLS and Segment Routing Operation

Segment Routing Operation	MPLS Operation
Segment Routing Header/Segment List	Label Stack Header
Active Segment	Topmost Label
Push	Label Push
Next	PoP
Continue	Swap
Segment Id	MPLS Label

There are two main types of segments in Segment Routing:-
1. Node Segment (Node/Prefix SID):- Node segment ID which is also called as Prefix segment ID is used to specify loopback interface of Segment Routing enabled device. The forwarding is associated with Node segment ID. The operator assigns a domain wide unique Node segment ID for each router in the network. This can be done manually or using a centralized controller (SDN Use Case).



2. Adjacency Segment:- Each router will assign a locally significant segment ID for each of its IGP adjacencies and it is not globally unique like Node/Prefix SID. Segment Routing enabled routers allocate Adjacency segment ID for their all attached interfaces automatically when the segment routing is enabled on the router.

How to make network ready for SDN/NFV?

Since long, Telecom Service Providers are procuring network equipment and appliances – such as core, edge, aggregation routers and switches. But new approach is required for these processes have to be modified to exploit the flexibility, agility, faster go to market and cost-savings in virtualizing these functions. Operators are not going to replace all their existing networking gear and operational support systems in one step and replace them with software running on top of servers. At 30,000 the software approach looks very good but in fact it is too much expensive and risky. The migration to NFV will take in shorter steps wherein at first step SP should build infrastructure for virtualization and add the required skill sets in the operation team to support it. The next step both the networks must co-exist and work together. In the final step, SP can remove the legacy stuff and complete migration can be done on SDN/NFV.

As I have already mentioned, at 30,000 feet it looks very intuitive and very good in presentation slides but the real challenge is totally different. We need to understand the different kind of networks, business processes and business requirements. Once this is done, need to understand how network agnostic software layer can be laid to support every function to achieve business objective. To deliver services end to end in a hybrid, multi-partner network of networks, you need a migration plan that uses best practices to ensure you are able to maintain service levels without degrading customer experience.

If you would like to make your network ready for SDN/NFV, you need to take the following steps:-
1. Add strong NMS
2. Start Migration of Physical Servers to Virtualized Servers
3. Add the automation layer to spin the new VMs, delete VMs as per the threshold defined
4. Building Overlay Networks
5. Start Adding open controllers to support network APIs
6. Merge Network and Server infrastructure in common software layer

As a result of this, you will get faster go to market, competitive, saves lot of capex and opex, get rid from the logistics and better customer experience.