Cat6000 and MPLS.

[yucheng-zhao]

Hi,

i'm wondering if MPLS is supported on cat6000/6500 family.if positive, what role can a cat6000 play in a MPLS network. can it serve as PE or P router? or both? is any special module gonna be required on cat6000 in order to fit itself in the MPLS network? or a simple IOS upgrade will be enough to do the job?

please help. thanx in advance!

[xavierchang]

I hate to answer a question with another question, but why do you want to implement MPLS at all? MPLS makes your network significantly more complex to manage. Don't believe the hype from the vendors that you must move to MPLS or you will be obsolete - your network will work just fine for the foreseeable future without it (a non-MPLS network will probably work even better than an MPLS-enabled network because MPLS is still not totally mature). You should have very specific and well investigated reasons for implementing MPLS.

This goes double for implementing MPLS-VPN's (I assume you are thinking about VPN's because you talk about PE routers). After all, let's just say that right now there isn't exactly a huge outcry of customer demand for MPLS-VPN connectivity. So it's not as if you can enable your network and your sales department's phone will then be ringing off the hook constantly from new customers. Right now, seems to be a lot of work in upgrading a network for miniscule payoff.

But in answer to your direct question, a Cat6000 can indeed be MPLS enabled with the right IOS image, and can be both a PE and a P router. No, you don't need any special modules. A simple upgrade will do.

But again, I must ask, why?

[yucheng-zhao]

LOL...

First off, i wanna thank you for your detailed answers to my question. really appreciated it.

and as for the question you brought up: Why the hell do we need a MPLS network anyway?!Geez, believe it or not,i've been asking myself the same question all the time,too. and i have to say: i couldn't agree with you more.

MPLS makes the current core of Internet more complex and hard to configure and manage, it makes the core more intelligent but the core DOESN'T need to be intelligent. High speed forwarding, that's what the core need. MPLS dose NOT scale. besides, there is no proof that label switching is faster than normal switching process.

and MPLS VPN is not even secure at all. one mis-configured command might leak the routing info from one VPN to others. no evidence has showed that MPLS VPN is more secure than IPSec VPN.

MPLS is not a perfect technology as most people think it is. but sadly, it IS the most-talked about technology, that's why i got that cat6000 question. anyway, it doesn't hurt to know that cat6000 is MPLS-enable, right?

:)

[xavierchang]

To add a bit to your analysis:

MPLS does indeed by itself more complexity to the core. Yet this can be somewhat alleviated by the ability to run a BGP-less core, so the net effect will be to significantly reduce the complexity of your core (BGP is significantly more complex than LDP/TDP or RSVP-TE). Yet I doubt that ISP's will adopt the BGP-less core model soon, if ever, because the fact is ISP's are loathe to give up the control that BGP offers.

On the other hand, MPLS radically increases the complexity and load of your edge, especially with VPN's (L2 or L3), and since your edge routers are typically not your strongest routers, this does not bode well at all. This is particularly true if you are thinking of packaging Internet connectivity along with your VPN's. Concepts like carrier-supporting-carrier can alleviate some of the worst scaling problems, but scaling is still a serious problem for the edge.

There is indeed evidence that label-switching is faster than regular switching, especially in the old days before the latest ASICs. Nowadays, label-switching is still faster, but only by a minor amount, certainly not enough to justify the increase in complexity in your network.

The security aspect of MPLS VPN's is somewhat of a misnomer. It is no less secure than frame-relay or ATM, but you never hear about security complaints about FR/ATM. But speaking of that, I fully expect that RFC 2547-style L3VPN's will be greatly overshadowed by kompella or martini L2VPN's (especially kompella), again, because of scalability issues with RFC 2547.

But, again, perhaps the best MPLS move of all is not to do it until it becomes more mature. Like I said, customers aren't exactly champing at the bit to receive MPLS services anyway, so why rush into it?

[yucheng-zhao]

Thanx for the insights about MPLS/VPN. One thing i need to clarify is: as far as i know, there is a huge market for MPLS and MPLS VPN here in China. Telecoms are widly deploying MPLS VPN in MAN, plus, they are even considering implementing MPLS TE, and this trend tends to become overwhelming.

BTW, can i get a hold of you by the email address:xavierchang@hotmail.com? would love to keep in touch. :)

[xavierchang]

Sure, you can email me anytime, and I'll be happy to tell you how MPLS works, or more likely, doesn't work.

Hmmm. I have no idea why Chinese telcos would want to jump fullbore into a technology that has gained only mild acceptance in North America and Europe (for the reasons specified before, namely lots of complexity with limited benefit). Is there some kind of huge groundswell of demand for VPN's in China that I am simply not aware of?

Actually, I consider MPLS TE, while still rather niche, to be significantly more practical in the short-term than VPN's. Right now, providers are fixated on cutting costs, and TE actually offers some cost-cutting possibilities, although not a huge amount because the simple fact is that most providers have a ton of excess capacity and don't have to concern themselves too much with using TE to use that capacity efficiently. But still, a somewhat decent business case could still be made for TE, at least much more so than the case for revenue-generating services like VPN's which as far as I can tell, customer demand is negligible (unless, I guess possibly in China, but I'm still not convinced of that).