x
Optical/IP

Marconi Adds 'c' to Its OC192

In today's competitive landscape, vendors are slugging it out for every available carrier dollar. So it's not surprising to see an increase in force behind claims about product features and customer wins.

Marconi plc (Nasdaq/London: MONI) is the latest example. The vendor says its recently unveiled BXR-48000 (see Marconi Gets a Boost) is the first and only multiservice switch to run Asynchronous Transfer Mode (ATM) traffic over a fully concatenated OC192c Sonet spigot.

Clarification: Sonet OC192c is different from OC192 in a key respect. While both support 10-Gbit/s data rates, OC192c does so via a single channel, whereas plain old OC192 multiplexes four 2.5-Gbit/s channels together. One channel is easier to manage and scale than four glommed together, comprende?

Further clarification: What Marconi's claiming is not that it's the first to have OC192c networking channels. Router vendors like Cisco Systems Inc. (Nasdaq: CSCO) and Juniper Networks Inc. (Nasdaq: JNPR) already offer those. What's new is the ability of a switch to do so -- specifically a switch that's geared to supporting legacy and Internet Protocol (IP) traffic in a single box (for more on such products, see Multiservice Switches).

Marconi's backing up its claim with testimony from its first BXR-48000 customer, the U.S. Naval Research Laboratory (NRL), which also helped Marconi develop the product.

"We have access to a lot of products, and I haven't seen any that can do ATM over OC192c," says Hank Dardy, chief scientist at NRL. He says, thanks to a blade delivered by Marconi this week, his agency is running live large database traffic across six government sites. "I think it's a first at this level."

Marconi's BXR-48000s are linked in Dardy's network to core switches from Firstwave Secure Intelligent Optical Networks Inc. (see Firstwave Follows the Feds).

Is this really unique? So far, it appears to be. "If they're running ATM over OC192c, that's different," says David Passmore of Burton Group.

Is this really important? So far, it's not clear that having OC192c is as important in multiservice switches as it is in routers, where demand already has pushed supply. And other vendors are taking their time. A source at Équipe Communications Corp. says his equipment has OC192 capabilities but probably won't have OC192c until chips become available, about one year from now.

Interestingly, Marconi says it designed and made many of the chips for the new OC192c blade in-house.

Nortel Networks Corp. (NYSE/Toronto: NT) says its Passport 20000 won't offer OC192c until the first quarter of 2003. Cisco claims its MGX 8950 multiservice switch supports OC192c (see Cisco Makes Multiservice Move ). But while Cisco says "quite a few" customers are trialing that platform, none, as of press time, were willing to talk about it.

"Having OC192c in a multiservice switch may not be essential today," says Passmore, "but as access speeds pick up, it will become very important to carriers."

Of course, Marconi hasn't made its OC192c blade generally available, so it's tough to verify its claims beyond the testimony of NRL. What's more, Marconi's blade needs to be considered in light of its other capabilities and in a range of configurations.

— Mary Jander, Senior Editor, Light Reading
www.lightreading.com
Page 1 / 6   >   >>
edgecore 12/4/2012 | 9:43:25 PM
re: Marconi Adds 'c' to Its OC192 --------------
One channel is easier to manage and scale than four glommed together, comprende?
--------------

Don't you also pay a pretty big header tax if you have to deal with 16 OC12 signals or with 4 OC48 streams vs one OC192?

EC
broadbandboy 12/4/2012 | 9:43:22 PM
re: Marconi Adds 'c' to Its OC192 An interesting question would be, under what circumstances would you need an OC-192c ATM SAR?

Are their any edge devices that can drive ATM cells at that rate?

I bet it would be more usefull to aggregate OC-48s from the edge onto a high-speed 10 Gig network, in which case you would only need switch-switch OC-192c ATM trunks.

But I could be wrong, especially with unknown military applications that are graphics/visualization intensive.

BBboy
Belzebutt 12/4/2012 | 9:43:22 PM
re: Marconi Adds 'c' to Its OC192 Can it SAR at OC-192 or does it just switch cells at OC-192?
Belzebutt 12/4/2012 | 9:43:19 PM
re: Marconi Adds 'c' to Its OC192 Let's say if you have a few big routers aggregating into this monster switch, say 4 OC-48 POS ports, all funneled into an OC-192 ATM port. I don't even know if Marconi offers POS ports though, if all it does is ATM then it doesn't need to SAR.

Being a multiservice switch I would expect them to offer Ethernet/POS interfaces...
jamesbond 12/4/2012 | 9:43:18 PM
re: Marconi Adds 'c' to Its OC192 Let's say if you have a few big routers aggregating into this monster switch, say 4 OC-48 POS ports, all funneled into an OC-192 ATM port. I don't even know if Marconi offers POS ports though, if all it does is ATM then it doesn't need to SAR.

---------------------------------------

Belzebutt,

Lets say it did offer POS. Now how would the
switch know that it has to SAR the 0C-48
inputs and send it over POS interface? is
there a command that tells the switch that
this is a routing interface?

just trying to learn.



pedantic 12/4/2012 | 9:43:18 PM
re: Marconi Adds 'c' to Its OC192 First of all, SONET is an acronym, not a word. We don't use Mpls or Atm so let's not use Sonet. Call me pedantic if you want--I did.

Second, (apologies to all the SDH folks, but I'm going to use North American speak for the concatenation stuff) any STS-n signal is made up of n Time Division Multiplexed STS-1s. SONET allows the payloads of these STS-1s to be concatenated together to create pipes greater than ~50Mbps. The 'defined' contiguous concatenations (Virtual Concatenation needs a whole article) are STS-3c, STS-12c, STS-48c and STS-192c although some boxes support STS-Nc where N is a multiple of 3.

So, an OC-192 (notice that I use OC and STS interchangeably) carries 192 STS-1s. They may be structured as anything from 192 individual STS-1s to a single STS-192c. So the statement that 'plain old OC192 (should be OC-192--remember, I'm a stickler for detail) multiplexes four 2.5Gbit/s channels together' is not correct. An STS-192 can be built from any combination of STS-1s and STS-Ncs.

As far as the comment about header tax goes, if you're referring to SONET overhead, it's basically the same regardless of the size of the concatenated pipes. Concatenated payloads have marginally lower 'overhead' than STS-1s since there are fewer fixed stuff bytes per data byte in the mapping format (1 per 87 vs. 3 per 87) but it doesn't matter what the concatenation size is. So the 'usable' bandwidth of 16 STS-12cs is the same as 1 STS-192c.

If you're not talking about SONET overhead, then pretend I didn't say anything about header tax.
edgecore 12/4/2012 | 9:43:15 PM
re: Marconi Adds 'c' to Its OC192 Good comments...still learning...

So what is meant when people talk about OC192 clear channel?

Can you also comment/educate also around the topic of channelized streams, is this a way to multiplex on a router (i.e. GSR has an OC12 card that channels DS3's...)

EC
broadbandboy 12/4/2012 | 9:43:14 PM
re: Marconi Adds 'c' to Its OC192 Good post, but getting past all the pedantic acronym stuff, why is it that most switch/router vendors seem to deliver a non-concatenated version, whether its OC-48 or OC-192, first, with the concatenated cards usually being delivered later. The question is whether the delay is due to lack of suitable phy chips, or due to difficulties with packet processing, queuing and memory, tm, etc.

I suspect its the latter.

BBboy

pedantic 12/4/2012 | 9:43:13 PM
re: Marconi Adds 'c' to Its OC192 Two replys in one...more for your money.

First edgecore:

>So what is meant when people talk about OC192 >clear channel?

This is an OC-192c--you get the whole (almost 10Gbps) payload for one channel.

>Can you also comment/educate also around the >topic of channelized streams, is this a way to >multiplex on a router (i.e. GSR has an OC12 card >that channels DS3's...)

I can't claim to be an expert on Cisco gear but my guess is that this card maps 12 different DS3s (~44Mbps payload) each carrying their own stream of packets or cells into 12 different STS-1 payloads in an STS-12. These STS-1s can be cross-connected within the SONET network to 12 different destinations.

I don't know whether these are DS3 streams built by the GSR or DS3 I/Os or a bit of both.

Now BBboy...

>Good post, but getting past all the pedantic >acronym stuff, why is it that most switch/router >vendors seem to deliver a non-concatenated >version, whether its OC-48 or OC-192, first, >with the concatenated cards usually being >delivered later. The question is whether the >delay is due to lack of suitable phy chips, or >due to difficulties with packet processing, >queuing and memory, tm, etc.

The PHY chips would be the same regardless of concatenation. It is only the devices that look at SONET pointers and process payloads that will care. Interestingly enough, for framers and data mappers, it can be simpler to support OC-48c/OC-192c, since there are fewer payload pointers to process and paths to monitor or generate (translates to less gates). I think the main issue is getting Network Processors to run fast enough to fill the big concatenated pipes.

Once again, my knowledge on the data side is a little thin, but my best information is that so called 10G Network Processors work well at 2.5G and if average packet sizes are large enough, they can fill an OC-192c. So, it would be a packet processing issue.
MP_UK 12/4/2012 | 9:43:10 PM
re: Marconi Adds 'c' to Its OC192 One channel is easier to manage and scale than four glommed together, comprende?
--------------

Don't you also pay a pretty big header tax if you have to deal with 16 OC12 signals or with 4 OC48 streams vs one OC192?
--------------

I think the reason concatenated payloads work better for ATM is that, if your framer chip gives you for instance, 1 10Gbps stream rather than 4 2.5Gbps streams, you can have connections above 2.5Gbps without doing IMA. Also you then only need a single connection to the switch core.
Page 1 / 6   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE