what is meaning of MPLS when compared to SD-WAN? is it dedicated MPLS or MPLS VPNs

Question

SD-WAN is always compared with MPLS, but it is not clear to me what is the meaning of MPLS in this context. My ambiguity is between these cases.

1. MPLS VPN services provided and managed by operators and service providers. I guess in this case there is not dedicated links but there is SLA.
2. MPLS links which are leased from service provider but is dedicated to the organization.
3. Dedicated MPLS links which are provided and organized by the organization itself and service provider has no role in this case.

I guess Cisco meaning of MPLS when compared to SD-WAN is case 2 or maybe case 3 (according to this link) but in some other documents their meaning of MPLS is MPLS VPNs which I think are not considered as dedicated links. Also, in some cases like this link I guess their meaning of VPN is application-layer VPNs not MPLS l2 and l3 VPNs. Can anyone tell me if there is real world WAN solutions for 3 cases above and explain why MPLS is generally compared to SD-WAN?

Answer 1

Can anyone tell me if there is real world WAN solutions for 3 cases above

There are many. Large organizations use MPLS in their own networks, especially when they provide "services" to other departments. In this case they act a bit like service providers, in the sense that they provide networking services to other groups within the organization. In other cases, there needs to be routing separation between various systems, and MPLS provides a convenient way to do that.

explain why MPLS is generally compared to SD-WAN?

SD-WAN is almost always presented as a lower-cost alternative to MPLS networks. SD-WAN (usually) uses lower cost Internet circuits instead of dedicated MPLS circuits^*. SD-WAN usually has some intelligence so that it can route time-sensitive traffic over MPLS circuits, and non-sensitive over an Internet VPN. In theory, you can provision much smaller bandwidth MPLS circuits,, since the majority of the traffic will use the Internet VPN. In practice, the savings aren't always so clear-cut.

*_{Note that MPLS uses dedicated circuits for the local subscriber connection, but the rest of the provider's infrastructure is shared. The ability to use shared infrastructure is the whole point of MPLS.}

Answer 2

MPLS is mostly used to send unencrypted data traffic over so-called label routes. This is done by a local infrastructure provider working together with several (if necessary) ASNs. This enables the infrastructure provider to do MPLS "switching" via BGP.

Some advantages of MPLS are certainly the high reliability of the connection between different company branches or data centers and the high quality (low latency and jitter). This is defined by SLAs and various contracts. These are binding and must be adhered to by the infrastructure provider.

However, MPLS has a very high price/Mbit. The more devices and the more data centers and branches are to be connected, the more expensive MPLS connectivity becomes. In addition, MPLS is not available everywhere and in every location. This means that company branches can often wait several months for a functioning MPLS connection until they go online or even receive it.

MPLS is, as already mentioned, not managed by the companies but by the infrastructure providers.

How an MPLS works:

How can an MPLS network meet such extremely good SLAs in the first place? MPLS is often described in the literature at OSI layer 2-3 or also assigned to OSI layer 2.5. The reason for the naming is how the respective MPLS router routes and switches the traffic.

At the entry of the infrastructure provider of an MPLS network, a label is attached to the data frame by the LER (Label Edge Router). The label contains the respective priority and the switching path, this is often also referred to as LSP or Label Switching Path. The LSR (Label Switching Router) forwards the packets to the next LSR. These only read the label and can read the LSP with it. The packets do not have to be segmented or fragmented. Through the egress router, the last LSR in the MPLS network, the packet is forwarded to the data center or company branch.

SD-WAN offers several options for connecting a company branch. It is always recommended to send data traffic going to data centers over a HA (High-Availability) line like MPLS, because as described above they have to provide certain performance due to SLAs and different contracts.

SD-WAN can therefore prioritize data traffic in the same way as MPLS. However, SD-WAN offers the advantage that data traffic can be routed directly to the public Internet through breakouts. This can be of great advantage when using any SaaS like Salesforce or Office365. There is no backhouling and depending on how many end devices communicate with SaaS, a lot of bandwidth is saved with MPLS.

As far as SD-WAN is concerned, there are many providers that exchange data traffic between one and another company branch in different ways. As an example, Cisco uses certificates and Sophos uses policy based routing via Sophos Central Servers. This means that an extremely healthy ecosystem can be addressed and with combination of MPLS, LTE, FTTH, (...) a redundant system of SaaS, Internet, Datacenter and Office/Branch connections can be built. This also means that company branches can be connected within a few days if they are often connected via SD-WAN, for example. The connection between the respective SD-WAN devices is encrypted and completely protected, even on the public Internet.

However, it must be noted that no guarantee of performance can be provided for breakouts to the public Internet via LTE, FTTH, etc. This also means that, even if HA is used, the company branch can be connected within a few days. This also means that even if HA exists, there are no SLAs.

On the whole, it is healthy to have a mixture of both to cover HA and performance.