MPLS Benefits???????

Answered Question
Feb 10th, 2007

Hi all,

I am investigating why we, as an enterprise, should move toward MPLS. What benefit does it give you? Anyone implements MPLS (not buying MPLS service) on their network and do you notice any difference? Do you gain any more speed? Are you able to control traffic better than tuning/redesigning IGP? Why QoS on MPLS is better? Please let me know. Thank you!

I have this problem too.
0 votes
Correct Answer by mikedavi1 about 9 years 7 months ago

Hi Kevin,

There are a few applications of MPLS for enterprise that may or may not have value for you.

1) VPN support. A la RFC2547, you may, as dvr0 suggested, segregate L3 information including address space, routing policy, reachability, etc. For example, you may have a subsidiary or acquisition that you provide transport for, but you want to keep separate in all other ways. However, you want to save costs by not buying additional hardware or circuits to support their traffic. MPLS L3VPN can be leveraged.

2) Layer 2 tunneling. Though there are several ways to accomplish this, you may want to extend the VPN concept by interconnecting and/or interworking layer 2 devices like LAN or ATM Switches, HDLC or Frame-relay, etc., across your existing infrastructure, without impacting other users and without buying dedicated circuits. AToM is one approach to accomplish this.

3) Traffic engineering. Suppose you want to take advantage of unused bandwidth, or determine the network path based on bandwidth, custom cost, or custom attributes such as time of day, affinity, traffic type, Int Serv signalling, etc,. The path chosen can be either dynamic or deterministic. And the solution needs to be scalable, manageable, and converge extremely quickly. MPLS Traffic Engineering is an approach that meets these requirements.

These are some of the most popular applications of MPLS and are all useful for and being adopted by enterprises.

HTH,

Mike

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4.5 (2 ratings)
Loading.
michelot2 Sun, 02/11/2007 - 12:09

Bonsoir Kevin,

> I am investigating why we, as an

> enterprise, should move toward MPLS.

MPLS is not done for LAN. It is used by carriers to interconnect different entreprise sites on a high rate backbone.

> What benefit does it give you? Anyone

> implements MPLS (not buying MPLS service)

> on their network and do you notice any

> difference?

The client buy a VPN service to the carrier and the carrier can use or not the MPLS to allocate a label to the client. There are other possibilities, but MPLS is today quasi universal.

> Do you gain any more speed?

In this architeture e.g Ethernet over MPLS over GFP over SDH over WDM, MPLS is a layer 2 procedure. So the speed is the speed of the physical layer protocol.

> Are you able to control traffic better

> than tuning/redesigning IGP? Why QoS on

> MPLS is better?

For that, you can see the ITU-T recommandations, in the Y serial. There is new recently information on that subject. But, once more, it is the problem of the carrier, not of the private entreprise that want to transport data to remote sites. MPLS is transparent to the client that buy a service with a SLA.

Best regards,

Michelot

Darren Ramsey Sun, 02/11/2007 - 16:00

Kevin,

Cisco classifies MPLS as an emerging technology for Enterprise Networking. It can be used for virtualizing and segmenting campus networks into multiple virtual private networks, provided you have the right hardware to run LDP and VRF. We use MPLS in the Campus to segment our guest network, and to resale bandwidth to physician practices located at multiple sites. We can provide say 100MB much cheaper than MetroE carriers can. Additionally we bring their traffic back to a firewall and can permit limited access to each other's resources (say Pacs or HL7 feeds) via policy.

Below are several docs on Campus MPLS.

http://www.cisco.com/web/strategy/docs/healthcare/MGN_Architecture.pdf page 23

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns251/c649/ccmigration_09186a0080680606.pdf page 8

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns656/c649/cdccont_0900aecd804ab672.pdf page 80

miclacs13 Mon, 02/12/2007 - 01:57

Hi,

Simply because MPLS VPN is the efficient and cost effective deployment of your end to end IP network. It could be cheaper than your traditional leased lines and it doesnt depend on what L2 access do you have per site.

Since you connect to the ISP directly by IP, you can demand and fall under specific SLAs. It makes easier also for telcos to manage including your WAN network.

Must take note that customer must be keen in policing the telco to take care of your COS/ QOS and vice versa that depends on your SLAs including attributes like congestion, latency and etc.

HTH...

Br,

Mike

michelot2 Mon, 02/12/2007 - 03:49

Bonjour Mike,

> It could be cheaper than your traditional

> leased lines

Don't forget that MPLS VPN are almost entirely transported on TDM networks (leased lines).

Best regards,

Michelot

miclacs13 Tue, 02/13/2007 - 17:49

Hello Michelot,

Yup, it could be delivered on any Layer 2 access be it leased line, frame relay, dsl or metro-E etc...

But in the sense of delivering global end-to-end MPLS circuit compared to end-to-end point to point international leased lines, its much cheaper. :)

Br,

Mike

michelot2 Fri, 02/16/2007 - 10:40

Bonsoir Mike,

> Yup, it could be delivered on any Layer 2

> access be it leased line, frame relay, dsl

> or metro-E etc...

We have to distinguish the transport MPLS (T-MPLS specified bu ITU-T) for which the carrier offers trusted VPN (not confidential VPN) from the MPLS specified by the IETF.

In WAN (and not in the local loop) Ethernet is often carried over T-MPLS and over a TDM server layer (e.g. VC-n-Xv through the GFP procedure).

> But in the sense of delivering global

> end-to-end MPLS circuit compared to

> end-to-end point to point international

> leased lines, its much cheaper. :)

From a technical point of view, it's too much ambiguous. The T-MPLS coud be also end-to-end, from PE to PE in a provider network.

Whatever MPLS or T-MPLS, data are carried over leased lines in long range networks.

Best regards,

Michelot

guruprasadr Wed, 02/14/2007 - 01:56

HI Kevin,

Major Benefits of MPLS Network is: It will summarize the Subnets inside the Network & will advertise only the Major Subnets (Global Subnets) between the WAN Links.

Inorder to avoid the Router devices to keep processing lot of Subnets in its Routing table, we use MPLS Network to have only Major outgoing Subnets in the Routing Table.

By, this Technology the Router Processor Load reduces totally in processing only the Global Subnets.

If helps please Rate the Answers !!

Thanks & Regards,

Guru Prasad R

kevin.hu Wed, 02/14/2007 - 06:40

Thanks Guru,

From what you are saying, MPLS reduces routing table due to the summarization, then how is it different, say from OSPF, where I can do the same thing on the ABR, or say from ASBR, I can summarize the redistribution subnets too?

guruprasadr Wed, 02/14/2007 - 21:03

HI Kevin,

Consider OSPF,

Summarisation between different AS is not possible. Each AS will have backbone Area 0 & and other Area as 1, 2 or 3 even more.

You need some Border Protocol to Communicate between different AS.

Consider BGP,

Route Summarisation between Different AS is possible. Use E-BGP between routers of different AS & Use I-BGP between routers of same AS.

Rate this Answer if Helps ! ! !

Thanks & Regards,

Guru Prasad R

Correct Answer
mikedavi1 Fri, 02/16/2007 - 19:39

Hi Kevin,

There are a few applications of MPLS for enterprise that may or may not have value for you.

1) VPN support. A la RFC2547, you may, as dvr0 suggested, segregate L3 information including address space, routing policy, reachability, etc. For example, you may have a subsidiary or acquisition that you provide transport for, but you want to keep separate in all other ways. However, you want to save costs by not buying additional hardware or circuits to support their traffic. MPLS L3VPN can be leveraged.

2) Layer 2 tunneling. Though there are several ways to accomplish this, you may want to extend the VPN concept by interconnecting and/or interworking layer 2 devices like LAN or ATM Switches, HDLC or Frame-relay, etc., across your existing infrastructure, without impacting other users and without buying dedicated circuits. AToM is one approach to accomplish this.

3) Traffic engineering. Suppose you want to take advantage of unused bandwidth, or determine the network path based on bandwidth, custom cost, or custom attributes such as time of day, affinity, traffic type, Int Serv signalling, etc,. The path chosen can be either dynamic or deterministic. And the solution needs to be scalable, manageable, and converge extremely quickly. MPLS Traffic Engineering is an approach that meets these requirements.

These are some of the most popular applications of MPLS and are all useful for and being adopted by enterprises.

HTH,

Mike

michelot2 Sat, 02/17/2007 - 09:51

Bonsoir Mike,

> For example, you may have a subsidiary or

> acquisition that you provide transport for,

> but you want to keep separate in all other

> ways. However, you want to save costs by

> not buying additional hardware or circuits

> to support their traffic. MPLS L3VPN can be

> leveraged.

Difficult to see the real technology behind this commercial description which can be the same whatever the context.

Could you please show a stack of protocols?

Thanks,

best regards,

Michelot

mikedavi1 Sat, 02/17/2007 - 15:37

Bonsoir Michel,

In the end analysis, is it not for commercial reasons that many 'carriers' choose MPLS?

In you post earlier, you say.

"MPLS is not done for LAN. It is used by carriers to interconnect different entreprise sites on a high rate backbone."...

... "But, once more, it is the problem of the carrier, not of the private entreprise that want to transport data to remote sites. MPLS is transparent to the client that buy a service with a SLA."

I would challenge those statements. In many ways the role of "carrier" is not different than the role of a networking department in enterprise. Both solve the same sorts of problems and provide the same services. For example:

Both work to provide transport for enterprise data.

Both try to find efficiencies in terms of OAM&P.

They both may provide transport infrastructure.

They both may provide SLAs.

They both may sell transport services.

They both may provide application services.

They both must manage cost. Cost of wire/fiber, cost of equipment, cost of personnel, etc.

And both solve the same sorts of 'customer' issues, but the enterprise will often have a greater responsibility for some issues than the telephone company.

They both may have customers with overlapping address spaces that must not intermingle.

They both may provide SLAs for 'customer' traffic.

They both may provide some security via traffic isolation.

My point in my earlier post was that enterprises can face several of the same challenges that carriers do. And just like traditional telephone companies have done, enterprises may choose to address those challenges with the same technology that many telephone companies chose.

To directly answer Kevin's first questions:

Do you gain any more speed?

- It depends.

-If the path selection control allows you to leverage unused paths, then overall, throughput increases.

However, label switching doesn't inherently make traffic move faster.

"Are you able to control traffic better than tuning/redesigning IGP?"

- Yes! This is MPLS's major strength. The extent to which this is true is what allows MPLS VPN and Traffic Engineering applications to be possible.

"Why QoS on MPLS is better?"

Better in some ways, worse in others.

A Benefit: Greater control of traffic path, independent of routing protocol. Ability to make best path selection based on available bandwidth. (Traffic Engineering)

A Drawback: Fewer bits to mark traffic with means less traffic classifications available.

Anyway, as you requested, here's some stack information for some relevant protocols:

MPLS

LDP

RSVP + TE extentions

BGP4 + IGP extensions

OSPF with Type 10 (Opaque) LSA support

BGP with VPNV4 address family support support

AToM - L2 Pseudowire support

MPLS reading list:

http://www.ietf.org/rfc/rfc2370.txt - OSPF Opaque LSAs

http://www.ietf.org/rfc/rfc3031.txt - MPLS Architecture

http://www.ietf.org/rfc/rfc3032.txt - MPLS Label Stack Encoding

http://www.ietf.org/rfc/rfc3036.txt - LDP

http://www.ietf.org/rfc/rfc2547.txt - MPLS VPN

http://www.ietf.org/rfc/rfc2205.txt - RSVP

http://www.ietf.org/rfc/rfc3209.txt - RSVP + TE extensions

http://tools.ietf.org/html/draft-martini-l2circuit-trans-mpls-00 - AToM

HTH clarify,

Mike

kevin.hu Tue, 02/20/2007 - 07:38

Hi Mike,

Thanks. These info are extremely helpful. So, currently we have point to point circuits, eg. T3, OC-12, T1 etc. Are we able to migrate these circuits toward L2VPN for redundancy reason (subscribe MPLS L2VPN service) and then on top of that running L3VPN so we can control L3 traffic ourselves? Is it possible to run L2 and L3VPN together?

mikedavi1 Tue, 02/20/2007 - 12:25

Glad to help.

Breaking your query into a few more manageable questions:

1) Are we able to migrate these circuits toward L2VPN for redundancy reason (subscribe MPLS L2VPN service)?

-Potentially yes, however your circuit provider will need support this. The nature of L2VPN is such that you may already be traversing a L2VPN carrier, but you as subscriber, will not/should not be able to tell. The service will/should be transparent to you meaning you'll just see a point to point T1, T3, etc.

2) "and then on top of that running L3VPN so we can control L3 traffic ourselves? "

-Yes. I have worked with a number of customers that wanted carrier diversity without the difficulty of finding two Inter-AS capable providers willing and able to peer, nor the potential re-provisioning of routing, IP addresses, etc. if a carrier's services were replaced. You can buy L2 transport from any carrier, and overlay (frame-mode) mpls or rsvp+TE extensions over the L2 transport and do the L3 yourself. Furthermore, you can overlay your own L2VPN transport over your MPLS overlay, all riding on the carrier's transport or your own.

3) Is it possible to run L2 and L3VPN together?

-Yes.

Mike

guruprasadr Mon, 02/19/2007 - 01:54

Hi Kevin,

Border Gateway Protocol v4 is an Exterior Gateway Protocol type (versus OSPF for example which is an Interior Gateway Protocol type).

1 BGP is made to connect IGPs between them and RFCs advise against OSPF to OSPF redistributions

2 BGP provide lot of attributes to manage the routing policy

3 BGP is a loop free protocol (versus OSPF to OSPF redistributions)

4 BGP provide routing interoperability between different equipments (OSPF to OSPF is Cisco proprietary)

So BGP has been choosed to provide interoperability, a robust, loop free and standard routing protocol over the Alcanet international backbone.

BGP basics

BGP use TCP connections to establish a routing path between 2 routers. This is so a point-to-point protocol versus OSPF, RIP or EIGRP for examples which are using broadcast or multicast to establish a routing neighboring.

There are 2 types of BGP peers :

1. External BGP peers, which provide a routing path between 2 routers belonging to 2 different Autonomous Systems

2. Internal BGP peers, which are mandatory to provide a BGP routing path between 2 routers belonging to the same AS each of one possessing some EBGP sessions

BGP is a path-vector (sometime called distance-vector) protocol with enhancements as reliable updates, triggered updates only and rich metrics (called path attributes).

Cisco BGP complete decision process :

1. Only consider paths with reachable NEXT_HOPs

2. Do not consider iBGP path if not synchronized

3. Highest WEIGHT

4. Highest LOCAL_PREF

5. Prefer locally originated route

6. Shortest AS_PATH

7. Lowest ORIGIN code: IGP < EGP < incomplete

8. Lowest Multi-Exit Discriminator (MED)

8a. If BGP deterministic-med, order the paths before comparing

8b. If BGP always-compare-med, then compare it for all paths

8c. Considered only if paths are from the same neighbor AS

9. Prefer an External path over an Internal one

10. Lowest IGP metric to the NEXT_HOP

11. If multipath is enabled, the router may install up to N parallel paths in the routing table

12. For E-BGP paths, select the "oldest" to minimize route-flap

13. Lowest Router-ID Originator-ID is considered for reflected routes

14. Shortest Cluster-List Client must be aware of RR attributes !

15. Lowest neighbor IP address

Administrative distances

Route Source Default Distance

Connected interface 0

Static route 1

Enhanced IGRP summary route 5

External BGP 20

Internal Enhanced IGRP 90

IGRP 100

OSPF 110

IS-IS 115

RIP 120

EGP 140

Internal BGP 200

Unknown 255

Rate this if Helps you in Understanding !!

Best Regards,

Guru Prasad R

Actions

This Discussion