Marconi Adds 'c' to Its OC192

COMMENTS

Newest First | Oldest First | Threaded View

<< < Page 6 / 6

jamesbond 12/4/2012 | 9:42:05 PM

re: Marconi Adds 'c' to Its OC192 Size - the larger the cell the less choping and fewer steps in reassembly.

>> really? 53 to 64 makes that much difference?

I want to ask a question though - do routers really re-assmble IP packets to chop and send over a fabric?? It would seem very inefficient to receive 53 byte cells, convert to IP frames - somewhere between 40 bytes and 1526 bytes (MPLS with standard LLC Snap packets) just to chop again into 64 bytes to rassemble and ship.

>> I hope you are not talking about IP
fragmentation and re-assembly here. Assuming
that yes it is inefficient. Maybe that is why
in general routers don't have ATM ports.

If so - it is very clear why routers don't manage ATM switching well.

Reply | Post Message | MESSAGES LIST | START A BOARD

testsweet 12/4/2012 | 9:42:00 PM

re: Marconi Adds 'c' to Its OC192 NM,

I am interested in your comments. If you wanted to chat offline you could reach me at [email protected]

Cheers

Reply | Post Message | MESSAGES LIST | START A BOARD

broadbandboy 12/4/2012 | 9:41:53 PM

re: Marconi Adds 'c' to Its OC192 ragnar_d_pirate wrote: "His name was Mahesh Ganmuki (sp?). After he left Ascend, he went on to form Ignitus which was later consumed by Lucent. After that he was CEO at Cereva, which is now defunct - I haven't heard anything of him since."
-------------------------------------

Thanks Ragnar, I found some info on him. Looks like he had a good track record until the last one. Maybe he should go back to working on the next gen of ATM switches...

>>Mahesh N. Ganmukhi joined Cereva Networks Inc. as President and Chief Executive Officer in May of 2001. Most recently, he served as vice president and general manager of ISTN at Lucent. Prior to Lucent, he was the founder, president and CEO of Ignitus Communications, Acton, MA which was acquired by Lucent in April of 2000. Before Ignitus, Mr. Ganmukhi was Director of Technology for the Core Switching Division of Ascend, Westford, MA, where he was the lead architect for carrier-class ATM switches. Prior to Cascade, he developed switch architecture and managed the ASIC development group at FORE Systems. Ganmukhi also developed Hypercube and Fat-Tree based network ASICs for massively parallel processors at Thinking Machines Corporation, Cambridge, MA.

This could be the beginning of a new "Where are They Now" series for LR...

BBboy

Reply | Post Message | MESSAGES LIST | START A BOARD

Belzebutt 12/4/2012 | 9:41:48 PM

re: Marconi Adds 'c' to Its OC192 I want to ask a question though - do routers really re-assmble IP packets to chop and send over a fabric?? It would seem very inefficient to receive 53 byte cells, convert to IP frames - somewhere between 40 bytes and 1526 bytes (MPLS with standard LLC Snap packets) just to chop again into 64 bytes to rassemble and ship.

That is exactly what they do. They reassemble the ATM cells into IP packets (SAR), do a routing table lookup, then chop the packets into proprietary cells, switch them across the fabric, reassemble into packets, and then send them out the egress port. If the egress port is an ATM port, that means chop the packets again into ATM cells.

This is inefficient but that's what routers have to do with ATM traffic: they route, i.e. they look at the IP header, so they must reassemble the entire packet.

Obviously an ATM switch will be better at switching plain ATM cells. A router's ATM interface is only really used as an access interface between your IP and ATM network, it's inefficient and more expensive to route between ATM interfaces on routers.

POS interfaces are much better for routers because they can transmit entire packets without chopping them into cells.

It all depends if you have an ATM network or an IP network in the first place, and if your customers want ATM PVC's or just access to an IP network. Or maybe you lease the connections between your routers from an ATM service provider. Many possibilities.

Reply | Post Message | MESSAGES LIST | START A BOARD

ajalgujal 12/4/2012 | 9:41:32 PM

re: Marconi Adds 'c' to Its OC192
Routers use proprietary cells - 64 byte cells, to switch across fabric

ATM switches use 53 byte ATM cells.

if routers have OC192 POS ports which can operate at line rate, why they don't provide OC192 ATM ports?. Anyway they have capability to chop the IP packets to 64 byte cells and switch. can't the chip do 53 bytes also?

Reply | Post Message | MESSAGES LIST | START A BOARD

Belzebutt 12/4/2012 | 9:41:26 PM

re: Marconi Adds 'c' to Its OC192 I'm no router switch fabric designer, but I would guess that it's just a lot easier to chop IP packets into simple proprietary cells than to segement packets for ATM adaptation layers. While the two processes would seem similar on the surface they are probably very different when you get down to the details.

Where's Tony when you need him?

Reply | Post Message | MESSAGES LIST | START A BOARD

TheNet 12/4/2012 | 9:41:01 PM

re: Marconi Adds 'c' to Its OC192 The problem is not such much about segmenting packets. It has more to do with reassembling the packets.

The reassembly process needs to queue separately cells of all VCs that require reassembling. Then, it needs to go through the list of cells in the queue of a VC and determine if it can reassemble a packet.

In short: lots of context & memory is required.

Reply | Post Message | MESSAGES LIST | START A BOARD

dlazar 12/4/2012 | 9:30:59 PM

re: Marconi Adds 'c' to Its OC192 As you already know, The Tolly Group enjoys a unique position in the networking industry. As the premier independent test lab, it has worked with virtually every major vendor and evaluated the most important products and technologies to appear over the past decade (including ATM).

This working knowledge of products and technologies puts The Tolly Group in a class by itself. Its strategic recommendations are always based on firsthand experience, rather than abstract trend analysis or simple customer surveys.

Both businesses and government agencies turn to The Tolly Group first when they need to assess existing and emerging technologies.

The Tolly Group has tested ATM equiptment from FORE, IBM, Lucent, NetTest, SUN, Marconi, and others and has the expertise and knowldege base to continue testing ATM.

Reply | Post Message | MESSAGES LIST | START A BOARD

<< < Page 6 / 6