Poll: Is MPLS BS?

A large proportion of the optical networking community takes the view that the development of MPLS (multiprotocol label switching) has lost its way, according to the results of Light Reading’s August research poll: MPLS - Just Kidding?

The poll, which has already been taken by more than 650 people, reflects a widespread belief that development work has been hijacked by the same folk that jumped on the ATM (asynchronous transfer mode) marketing bandwagon several years ago.

In particular, they say, MPLS is being promoted in the same way as ATM was -- as a panacea for simplifying networks and addressing every service provider requirement under the sun. Some network architects feel it’s being pressed into use for applications for which it isn’t well suited, like virtual private networks. (It does, however, make perfect toast.)

The upshot of this "over-egging the pudding" is likely to be over-complicated software -- and that translates into bugs and unacceptable reliability risks for service providers.

In our poll, 36 percent of respondents think that “MPLS is effectively ATM all over again,” and another 32 percent think there’s some truth in the statement. Only 28 percent disagree.

In some respects this isn’t particularly surprising, because the fundamental goal of MPLS is much the same as ATM's -- to enable service providers to consolidate traffic onto common infrastructure and offer predictable service quality .

A considerable proportion of respondents -- 38 percent -- think that “MPLS will end up just like ATM, a big ball of hair." Another 16 percent think there’s some truth in the statement. However, 40 percent flat disagree. (Could they be the folk marketing MPLS routers and switches?)

The split reflects a “religious” separation between two camps in the optical networking industry, according to David Drury, president of the MPLS Forum -- who just happens to have been a big ATM marketing muckamuck himself, at Fore Systems, the ATM switch vendor acquired by Marconi Communications PLC (Nasdaq/London: MONI).

Drury belongs to the camp that believes some services (such as voice and video) have to be provided over some sort of connection, whether it’s physical (as in a telephone line) or logical (as in frame relay, ATM, or now MPLS) in order to guarantee quality.

The other camp comprises "people that think everything can be done with connectionless protocols [like IP],” says Drury. “But no one has shown me how this can guarantee predictable quality services."

This second camp often takes the view that the easiest way to make service quality predictable is to over-provision bandwidth, particularly as optical technologies are driving down the cost of doing this. That’s what’s behind the final question in our poll, whether QoS is only needed at the edge of networks and not in the core. 50 percent of respondents believe this, and 33 percent don't. “In the wide area,” Drury concedes, "over-provisioning the core may well be okay."

To add your egg to the pudding, click here: MPLS - Just Kidding?

— Peter Heywood, Founding Editor, Light Reading
Page 1 / 18   >   >>
hisles 12/4/2012 | 7:55:26 PM
re: Poll: Is MPLS BS? Just some of the key things(others will add more)

1) Scales VLAN's at the edge of the network and allows L2 across a L3 boundary(routed cores)-point and point to multipoint Transparent lan services(TLS)

2) Allow Frame relay like security(PVC scenario now referred to as LSP's) and end to end services for pure IP VPN's_and even non ip in the TLS Martini draft scenario

3) Will allow all these non revenue producing simpletons(like 360 etc etc etc etc etc) to interconnect their networks to compete with the big boys and monitor the performance of individual LSP's---this requires the right choice of provision and management etc-I won't plug the vendors at this point

Simple terms
Allows the whole ethernet everywhere story to scale into carrier networks and lets the unprofitable "build and they will come" boys to gang up on the incumbents and deliver guaranteed SLA's from a wholesale perspective to resale broadband private Broadband IP VPN's without the need for firewalls(although financial institutions will probably still buy firewall)

Overhyped ...more like the holygrail

wdog 12/4/2012 | 7:55:24 PM
re: Poll: Is MPLS BS? David Drury is right the Gǣholy warGǥ is basically the same as it has always been. The circuit camp versus the packet camp. The IP crowd screams bloody murder about the evils of circuits, but almost all of the goals of MLPS are around making connectionless layer three (costly) IP offer the same attributes, (security, predictable latency and predictable jitter) as a simple (cheap) circuit. Instead of spending millions of dollars on routers and MPLS switches to implement VPNs and manage IP flows over Ethernet, it makes a whole lot more sense for carriers to invest in new generation SONET TDM switches (like the Ciena CoreDirector) and invest in systems to improve service provisioning. On a per $/Megabit basis, circuit switches will always be cheaper than either routers or MPLS switches and improvements in the systems to provision circuits are a lot less challenging to engineer and deploy than MPLS. IGm sure the connectionless crowd will attack me with the same old tired claims that TDM wastes bandwidth and SONET isnGt optimized for data. But I contend that improvements in provisioning systems that can cut provisioning times and move a few steps closer to provisioning bandwidth on demand do away with most of this argument. And, they are far more realistic than MPLS/IP/Ethernet everywhere. And, orders of magnitude more cost effective.
w 12/4/2012 | 7:55:22 PM
re: Poll: Is MPLS BS? I understand the argument from the packet camp that TDM wastes bandwidth and SONET isn't optimized for data. But doesn't the over-provisioning alternative also waste bandwidth by its very definition (although I do hear some say say it is not over-provisioning but correct provisioning)? Has anyone done a study on how much over-provisioning would be required in order to create ATM like QOS?

Another question to wdog: you seem to assume that MPLS is incompatible with SONET. Is that right? I was under the impression that it was compatible and that people were looking at implementing MPLS over SONET.
wdog 12/4/2012 | 7:55:21 PM
re: Poll: Is MPLS BS? W,

MPLS is a very over-used, over-worked term.
There is a portion of "MPLS & GMPLS" that is dealing with signaling which is basically the mechanism to set up, (you guessed it) a circuit across the network. This portion of the the MPLS activity is workable with SONET. However, there is still mountains of standards work to be done and agreed upon before carriers can benefit from it.
jmd 12/4/2012 | 7:55:21 PM
re: Poll: Is MPLS BS? Is there a difference between Edge and Core requirements?
itsmeagain 12/4/2012 | 7:55:19 PM
re: Poll: Is MPLS BS? SONET IP what else do we need?
dietaryfiber 12/4/2012 | 7:55:18 PM
re: Poll: Is MPLS BS? I think the problems MPLS will face will come from the implementation of QoS that is desired. I found it interesting that "ATM is a real problem", when in fact it is used in large scale networks all over the world. Where it gets messy is that ATM is optimized for fine grained QoS. If this is attempted by MPLS, then MPLS will have the same complication as ATM.

So, are there alternatives? Sure, but most of them are the kind of comprimizes that the IP and MPLS folks have said are not required. Undersubscribing bandwidth is similar to the ignoring of the loss of bandwidth to SONET framing or the loss of bandwidth to ATM "cellizing". Undersubscription keeps things to the simple QoS model of TDM (or ATM CBR). In fact ATM can be very simple if all that is desired is CBR and UBR.

dietary fiber
wdog 12/4/2012 | 7:55:17 PM
re: Poll: Is MPLS BS? dietaryfiber,

I'm not sure I understand everything you are saying, but one of the failings of ATM was it was attempting solve every problem known to the voice and data world. Where it ended up being very successful was as a traffic engineering solution allowing bandwidth to be provisioned in service compatible increments via PVCs. All of the promisies of ABR, VBR, UBR and SVCs never happened. The same traffic engineering solution can be done without the complexity of ATM protocols with the new generation of SONET.
wchieh6 12/4/2012 | 7:55:16 PM
re: Poll: Is MPLS BS? As I know, data aware SONET and Ethernet compete with each other in the metro market. Most of Ethernet camp use MPLS as QoS mechanism. If MPLS isn't such a wonderful thing as they are telling about, will data aware SONET be the winner in the battle?
flanker 12/4/2012 | 7:55:14 PM
re: Poll: Is MPLS BS? Peter:

I dont see a lot of support on this board for the article's position that MPLS is a non-starter. I agree it is in the early stages of development, but you havent even addressed the fact that most tier one carriers still run IP over ATM.

They would love nothing more than to replace that ATM gear with MPLS.

I know some very competent CTOs who are deploying greenfield networks and still refuse to set up the net without ATM protection.

What is the alternative? IP still needs QOS and policing.
Page 1 / 18   >   >>
Sign In