Zarlink Unveils Access Processor

A little chip announced by Canada's Zarlink Semiconductor Inc. (NYSE/Toronto: ZL) earlier this week highlights a new way of dealing with legacy traffic on packet-based networks (see Zarlink Set IP Networking Benchmark).

Zarlink says its chip, called an "access processor," can provision end-to-end TDM (time division multiplex) channels over a packet network. "Up until now it's only been possible to send voice data over an IP network," says a spokesperson. The new chip can provision up to 32 leased lines (T1/E1 channels) over IP, whatever kind of information they contain, and do so with carrier-class quality, the company claims.

In fact, the chip is a new take on the old idea of "everything-over-IP." It's based on a proposed standard called CESoPSN (circuit emulation services over packet-switched networks). This proposal, which was put forward by Axerra Networks Inc. is currently under consideration by the Pseudo Wire Emulation Edge to Edge (PWE3) working group at the Internet Engineering Task Force (IETF), whose mission is to find ways of emulating different protocols over an IP network.

It's pretty early days for this standard, but the introduction of Zarlink's access processor indicates that the company sees a reasonable market, reasonably soon, for products based on it.

So what's the big deal? According to Todd Shepherd, Axerra's director of systems engineering, CESoPSN is supposed to offer carriers a way of migrating their traffic onto a unified network infrastructure. Right now, carriers deal with multiple types of traffic in their networks by using overlays such as ATM or Frame Relay, he says. CESoPSN essentially collapses all the traffic onto a single packet-based network core, making the fact that it's travelled over a packet-based network invisible to the equipment on the ends of the link.

That's a big advantage to carriers and their customers, Shepherd asserts, because it means they don't have to rip out their existing TDM access equipment, as they would if they were moving to, say, Voice over IP (VOIP).

Jeremy Lewis, manager of Zarlink's TDM-to-IP product line, says one of the crucial differences of this approach is the ability to offer carrier quality on the links. The keys to this are timing and latency (delay).

"The timing of TDM streams is critical," says Lewis. "In TDM networks, the customer premise equipment takes its timing from the core of the network. If there is a packet network in the way, you can't do that any more." To overcome this obstacle, Zarlink's chip contains some clever algorithms to transport the timing information along with the payload as it travels over the packet-based core.

The chip also has low latency -- it introduces a delay of only 250 microseconds. Lewis contends that the only alternative to using Zarlink's chip is to buy an off-the-shelf RISC microprocessor, or digital-signal processor, and such a solution would have a delay of about two milliseconds -- eight times as long.

Of course, other equipment in the network, like routers and switches, could introduce long delays, which would screw up the latency (read quality) of the link. Lewis acknowledges that this could happen, but says network designers would bear this in mind.

"We're not talking about circuit emulation over the Internet," he says. A more likely scenario is a customer building a managed network, where they are trying to keep the overall latency to under 20 ms. Zarlink's chip makes it easier for them to stick to the latency budget. "If the devices at the end of the link introduce four or five milliseconds of delay, then that's half the latency blown already," he notes.

Zarlink claims to be testing its chip with "no more than five" alpha partners. The product hasn't been given a part number because it's the company's policy not to until it becomes generally available.

So far (and perhaps unsurprisingly, given its role in putting the standards proposal together), Axerra claims to be the only company to offer a system that delivers the benefits of circuit emulation (see Axerra Aims for Missing Link). It doesn't use the Zarlink processor; instead, it has built the technology into ASICs (application specific integrated circuits), and FPGAs (field-programmable gate arrays).

In fact, at least one other company, RAD Data Communications Ltd., has been offering "TDM-over-IP" equipment for several years. Its IPmux family of products supports up to 16 T1 (1.5 Mbit/s) or E1 (2.0 Mbit/s) lines over IP backbones. Swedish carriers were using RAD's boxes to offer low-cost leased lines over Gigabit Ethernet networks as long ago as 1999.

Other companies have similar things in the pipeline: Cisco Systems Inc. (Nasdaq: CSCO), Extreme Networks Inc. (Nasdaq: EXTR), and Nortel Networks Corp. (NYSE/Toronto: NT) are all involved with the PWE3 working group.

— Pauline Rigby, Senior Editor, Light Reading
Bumper_car 12/4/2012 | 10:31:13 PM
re: Zarlink Unveils Access Processor With the failure of broadband packet service providers like Yipes, where is the evidence that there is an end user market for TDM over packet networks? Packet networks are a lot more complex than TDM networks, and the packet network itself runs over TDM infrastructure. Since TDM services don't have any statistical gain, where services that provide TDM over packet be profitable?

Sure, service providers have been touting this new service or that new service for years, but it seems that the end users are buying the old traditional services and demanding traditional service reliability and stability. This may be the reason that service providers like Quest have backed off from even evaluating a lot of new technology. History has shown that creating a market need in transmission services dosn't work the way that it does with computer games and music.
Ether-Freak 12/4/2012 | 10:31:10 PM
re: Zarlink Unveils Access Processor User demand for scaleable data services in the access market will drive packet based mechanisms. TDM over packet is required to interface to legacy type voice PBX / Keysystems or to provide private line services that extend beyond the service providers domain where only TDM transport is available.

When service providers bring a MTU business related building onto their network via fiber optics, they do not want to provision multiple access systems. The business plan will drive the move to lower cost access mechanisms that will provide scaleable bandwidth for data while supporting existing customer legacy interfaces.