Going with MPLS network

[spidey07]

I'm in the process of re-doing our network. Everything.

One big pain point is the cost of frame-relay.

So after much financial, technical and strategy meetings with AT&T, sprint, MCI I've decided to replace/augment our frame with MPLS.

requirements from a WAN:
QoS - must support diffserv with at least 4 classes
Last mile - must support ranges from fractional DS1 to OC-3
management - must have "application visibility" similar to that already provided by frame-relay network on visual networks CSUs. Meaning we must be able to understand what applications are using the net and how much, deeper than simply utilization.
Scale - must be able to scale 50% per year. cost per Mb should decrease as it scales.
Security - must be as secure as layer 2 technologies like Frame.
Routing - convergence in under 30 seconds, support OSPF or BGP
Multicast - support for PIM

thoughts?
I'm wary of moving to it as frame has been great. But the pricing is good. My only concern is can it truly deliver the kind of quality of service I'm accustomed to with frame.

 

[cmetz]

I don't think MPLS is ready yet. It may NEVER be, because nearly the same folks who did ATM have reappeared doing MPLS, and in years and years and millions and millions the ATM people never did make something fully working. MPLS is just ATM all over again, really. Probably going to end up being just as big a CF, too.

It really depends on your needs vs. budget vs. reliability requirement. To get lots of bandwidth on the cheap, you trade reliability. What's most important to your enterprise? MPLS is the cool new technology the telcos are all herding over, and there don't seem to be many buyers as of yet. That probably means great deals to be had if you squeeze, but it also means you might be way out there from a support perspective (you don't want to be customer #5 of 10 for a network technology!).

Are you doing point-to-point or p2mp/MBNA? The former is probably a whole lot safer.

Augmenting sounds like a reasonable deployment strategy... bring up some links (hopefully parallel existing links) with MPLS and ease it in with some strategy to dump it if things go south.

Routing - ask pointed questions about the routing protocol underneath and the training their support folks have on it. One of the big big problems with smart L2 networks is that they all have their own special new routing protocols (newly implemented, typically ) underneath which the average NOC guy might not know how to debug. When some long line goes south and their MPLS edges re-signal and it doesn't quite work right, then what?

Multicast? Really?

 

[spidey07]

well Cmetz, you echo some of my concerns.

But at the same time when the frame is over 200K a month and I can go MPLS for the same bandwidth at 120K its simply too attractive to not try. (one of my key tasks/metrics is to reduce baseline costs)

But thanks a bunch for the comments. I took a gamble on frame 10 years ago, and thru the headaches it was the right decision.

We'll see. I'm working on a deployment schedule/migration plan.

I guess you can see my dilemma "provide more for less". Literally, less operating costs and at the same time traffic is increasing steadily at 50% year over year.

 

[AFB]

200k a month :Q

 

[ScottMac]

For a big enough account (I'm sure you are), you should be able to arrange a "pilot" type of contract: It would be multi-year contract with most of the heavy charges on the back-end. Add to that (as in negotiate) a fairly tight SLA.

If the service turns out to be flakey (repeatedly violating the SLA), then you have grounds to kill the contract, or at least get some of the money back based on unavailability / poor performance.

As far as the SLA goes, be careful about the measured time periods: most SLAs are written to cover a long term monitoring cycle .... such that a number of violations, when integrated into the entire monitoring cycle, remain below the violation level.

It's a relatively new service, they should be pretty flexible with the terms. Get the ugly charges back-ended; if the service turns out to be "not ready for prime-time" you can kill the contract and still get out cheap.

I disagree with CMetz about his take on ATM ... but it's an old / other discussion.

Next week I'll ask around over on the network side and see what they can offer on background details for the MPLS networks. If I get anything good, I'll pass it along.

Good Luck

Scott