Replacing Frame-Relay - with what?

[jroyster]

So Frame-Relay for a domestic WAN is becoming increasingly too expensive. Like a typical corporation all of our applications are housed in a Data Center in a hub and spoke fashion.

26 remotes office/manufacturing have frame-relay with ports ranging from 512 - 1544 Kbs.

And that is hardly enough bandwidth, circuits are still being overutilized.

So the question that plagues us is knowledge workers in these sites saying "well why come into the office and why are we paying so much for this circuit when I can go home to my 40 dollar a month Cable Modem and get 4-6 Megabits second. What is wrong with our net that it is so slow"

Slow being that a 15 Megabyte file doesn't instantly appear on their desktiop.

So what technologies should we replace frame with? Just go with DSL/Business cable service? MPLS? IP?

Even with talking to the carriers we're still limited by the last mile and 1544 Kbs. But I need multimegabit services...for cheap.

Any ideas or has anybody gone through similar pains?

[sschingel]

We had a similar situation where I work. We switched to a Point to Point T1 for every location. We installed a DS3 into our Data Center (channelized, it carries 28 T1's) and it seems much better than Frame Relay. I spoke with someone at SBC (our carrier) and FR goes thru a bunch of switches, so that's why ping times are longer than for PtP. I'm like you, we had some folks here complain that they can get stuff faster at home over their cable modem than they can at work. And this is true, to a degree, as I have a cable modem at home and it can be very fast. I have talked to telco's about speed, and it seems to be basically an issue of tariff's.

[scottmac]

When you buy frame-relay, you are also buying some flavor of Service Level Agreement (SLA). The additional cost of building a network that will support that SLA is part of the expense of a COMMERCIAL-GRADE circuit.

Home/consumer-grade cable or DSL circuits are delivered on a "best effort" basis. If the lines go down, that can be down for a while and not result in any serious bill-back or penalties.

You can buy (at least you used to be able to buy) "best effort" Frame-Relay as well: it was called "Zero CIR." It's not very practical for well-populated networks, since available bandwidth is fairly scarce and your throughput will be minimal.

Betting the business on "best effort" may work for some, but generally, these days, if the network is down, the business is down as well (or suffering some financial loss).

You can buy circuits emulating frame ... as mentioned above: you bring in a DS3 / Fractional DS3, and have the carrier aggregate the (one or more per site) T1s in the cloud and deliver them to the central site on a group of channels.

It's more expensive then Frame-Relay, because you now own 100% of that bandwidth, 24 x 7 x 365 instead of sharing the bandwidth as with Frame-Relay.

You can also (usually) bring in an IMA (Inverse Mux over ATM) circuit and build your bandwidth above traditional single-line levels.

Many Metro areas still have SMDS (Switched Multi-Megabit Data Service) with bandwidth generally hanging ~ six megabit.

Some areas (some SBC areas) have Gigabit point-to-point out to a 100 mile radius (SBC calls it GigaMAN).

There are a number of options amongst the variety of carriers ... but the common aspect to all of them is that to provide the quality and reliability demanded by most businesses, there has to be a significant investment by the carrier in infrastructure, monitoring, maintenance, support and planning.

What you will likely see in the not-too-distant future is significant bundling of products and services. By bundling, the provider can usually reduce the cost to the customer (overall), and share some of the expenses (like maintenance and support) among the various enterprise groups providing the individual services in the bundle.

Is is almost universally true that you get what you pay for: anyone that has suffered poor performance from "cheap" T1 providers (not all are bad, but many are) and slack SLAs, will understand why higher-quality service (with tighter SLAs) will nearly always cost more.

FWIW

Scott