5

I am studying about WANs and there's been a mention of several WAN options/technologies.

I know that Leased Lines are dedicated connections that are not shared by other customers while MPLS is a shared infrastructure that uses labels to keep different data separate and secure.

My question is, is there any major difference between Leased Lines/MPLS? Or better said, when would we want to use one over another? Are there any advantages/disadvantages?

I've also heard that Leased Lines are a thing of the past. Can someone please clear this for me?

1
  • The term "leased line" implies copper or service over copper. These days one can buy a point-to-point fibre circuit, where the fibreco is responsible for patching your two endpoints together either physically or logically. So while the phrase Leased Line is old, the concept is the same.
    – Criggie
    Commented Dec 7, 2022 at 22:43

6 Answers 6

10

One major difference between the two is that leased lines are generally more expensive than MPLS and can also be more reliable and offer higher bandwidth, because they are not shared.

MPLS on the other hand, is way more cost-effective option for connecting multiple locations, because it uses that shared infrastructure - It can also offer more flexibility and scalability.

In terms of when to use one over the other, it really depends on your specific needs and requirements. If you need a dedicated, high-bandwidth connection (or rather fixed bandwidth with no/low fluctuation), a leased line may be the better option. However, if you are looking for a more cost-effective and scalable solution, MPLS might be a better fit.

As for whether leased lines are a thing of the past, I wouldn't say so. While MPLS has definitely become more popular, leased lines are still commonly used by many businesses and organizations. It really depends on your specific needs and requirements.

0
4

A leased line offers a service where the bitstream that comes out is exactly the bitstream that comes in. Fixed throughput, fixed latency. There is also extremely limited latency added by intermediate equipment, as they can forward bit by bit.

MPLS services, like all packet or frame switching services, offers a service where the packets that come out are the packets that come in. But the end-to-end latency and the exact timing of the packets relative to each other may vary (jitter). In cases of contention somewhere in the network, there may be delays (or even packet loss). There is also slightly higher switching latency, as each switching/routing equipment on the way needs to read enough information from each packet to know where to forward it (depending on the implementation it may be just headers or it may be a full packet).

A leased line is analogous to a water pipe of a constant diameter. MPLS services more like a network of roads with vehicles on it, having to yield to other vehicles, possible traffic jams, etc.

For a point-to-point link a leased line is technically much better, but it's also a lot more expensive, as the network will have to reserve the exact capacity you buy from end to end on all links (there is no contention). Some of the links are just for you (usually the link from your premises to the first network equipment, and the same at the other end), some others are shared, but with a dedicated "lane" just for you (specific time slots or frequencies), even if you don't have anything to send on it. On packet or frame switched networks, they can mix traffic from many different source/destinations onto the same lane.

1
  • MPLS looks only at the label. The only time anything in the packet/payload is inspected is at ingress where the first label is built. That used to be why MPLS was superior to routing - label and LSP lookups are fast - today, hardware can do route lookups just as fast.
    – Ricky
    Commented Dec 7, 2022 at 23:47
2

Leased lines (also known as private wires) have been around since the days of the telegraph.

The same transmission technology is used in the local loop (and sometimes in the backhaul from the local exchange to the switch) for packet/frame/cell switched technologies (X.25, Frame Relay, SMDS, ATM) as in contemporary leased lines. Note : MPLS is not a transmission technology; a leased line could be changed to an MPLS link just by changing the software in the routers at each end. *1

With a long-distance leased line, resources are dedicated to the line in the telco backbone network (amplifiers and pairs in the early analogue days, timeslots in the PDH and SONET/SDH days). This requires more circuits and equipment than an MPLS network, where oversubscription of the bandwidth is usual. It was usual that the PDH data network was separate from the voice network, leading to further costs.

Therefore, telcos have been able to price MPLS and its predecessors more competitively than leased lines and have gradually phased out the latter. In the UK, Megastream services were still available for sale up to December 2021, but I doubt many circuits were sold for data in the 21st century (non-data leased lines still have their place in railway signalling and air traffic control, for instance).

Apart from the cost savings available, a company wishing to connect a number of sites together with leased lines either has to create a central hub with the capacity for one link per site (and take a risk that a failure of the central site shuts down the entire network) or lease extra lines to create a mesh.

With a packet/frame/cell/label switched network, the mesh is created by the router configuration and alternative routing is easily set up.This was an important reason in my personal experience managing a network of about 30 sites for replacing the multiple hub and spoke leased line network of 1993 with Frame Relay in 1998 and MPLS in 2003.

Another possibility offered by MPLS is the ability to mix public Internet access and private networks in the same local circuit and CE router. The CE router is not limited to one private network, so a multi-tenanted office could provide MPLS access for different tenants to their related sites with shared equipment. This requires a level of trust and bandwidth mix which seems unlikely to me.


*1 The introduction of fibre services changed the bandwidth available but not the nature of the services; long-haul Gigabit Ethernet was available on a leased-line basis in the early 2000s.

1

Just a quick note on leased lines; in the past when there was no fiber infrastructure, there were thousands of copper cables running between nodes and cities. As a company, for your critical data, you may choose to hire (lease) these copper lines from the provider between your offices. In my country, the provider would charge you monthly as if you make a call from one office to the other and did not hang up all month. By leasing a line from a circuit-switched network that is actually what you do. You block those two wires for your company. You may choose to use it as data or maybe a hardwired phone. The provider does not care since those two (or more) wires are yours. Greater the distance greater the price. Leased lines today's equivalent would be asking the provider to give you one (or maybe two) fiber cores to you(goes with the name dark fiber). If you calculate the potential of fiber cores today (including technologies like dwdm and Gpon), It would probably cost a ridiculous amount of money.

1

The true point-to-point lease line died decades ago.[*] A T1 between sites hasn't been an end-to-end copper circuit for a long, long time. (In fact, T1's haven't been "T1's" for decades. It's a 4-wire T1 from the smartjack to you, from the smartjack to the network it's single pair HDSL.) The T1 from each end goes back to their respective Central Office where they're are MUXed with other circuits onto larger circuits (T3, OC-3, etc.) between CO's. This involves rigidly mapping timeslots between circuits. Saying a private line is "not shared" is not entirely true.

MPLS does virtually the same thing... mixing traffic across larger circuits. MPLS just does it with a more modern flare. A T1 carrying an addressable encapsulation (frame-relay, or ATM) can be switched (MUXed) by virtual circuit (VC). (this was a common way to build point-to-multipoint networks.) MPLS can tag (label) traffic with much greater detail, not just the ingress interface, thus providing different paths for different types / classes of traffic. For example, voice is more sensitive to latency, that traffic is labeled for a path with the least delay -- which may not be the least hops.

[*] Dark fiber is still a thing, but there's a good chance it's being MUXed at some point as well. A literal bare fiber cross-connected between points would be impossible for the carrier to monitor.

3
  • mh, isn't the point of dark fiber that it's not muxed, so you can put whatever muxes you want there yourself? An also that the carrier doesn't and doesn't need to "monitor" what you do there.
    – ilkkachu
    Commented Dec 7, 2022 at 14:38
  • @ilkkachu: I hope "monitor" in this context means "notice if it fails", so they can start fixing it before you call them. Or monitor performance characteristics to notice pre-failure indicators. But yeah, I'd have expected "dark fiber" to have its literal meaning, not just some wavelength range on shared fiber. Commented Dec 7, 2022 at 16:10
  • @PeterCordes, well yeah, monitoring makes more sense. Though why bother, if you have multiple clients renting pairs from the same cable(s), probably one of them will tell you if there's a cut ;) Also as long as partial damage to a cable is rarer enough than full cuts, it's enough to monitor one or some of the connections on a given cable run.
    – ilkkachu
    Commented Dec 7, 2022 at 18:43
0

A single leased line or point to point fibre by itself isn't really a WAN - it is more of a network extension between exactly two sites.

If your needs are such that the two end points are your data centers, or a DC and a head-office, or between redundant replication sites then a dedicated link might be best.

By comparison, if your needs are to "connect all the sites" then a cloud can be more economical. It really depends what endpoints need to talk to what, with high/low bandwidth or latency requirements. Different sites have different limitations on their links too.

There are situations where you could have both, a MPLS cloud that learns routes advertised by branch sites, and a big fat pipe from the cloud to each data centers. The two DCs have a second separate direct link between them that is direct and dedicated, in case the cloud has issues.
Imagine a cloud outage that affects one DC - it would be possible for all branches to access those resources via the other DC then across the direct link.

Not the answer you're looking for? Browse other questions tagged or ask your own question.