I'm interested in peoples thoughts as to what is happening with the future of ATM NAPs. Are people moving away from them or are they still quite popular? Downsides or upsides? I'm trying to gauge whether they are worthy of investing equipment and hassle into for peering... -- Ben Buxton - Random Network Person
On Mon, 18 Dec 2000, Ben Buxton wrote:
I'm interested in peoples thoughts as to what is happening with the future of ATM NAPs. Are people moving away from them or are they still quite popular? Downsides or upsides? I'm trying to gauge whether they are worthy of investing equipment and hassle into for peering...
From a research university perspective, the AADS NAP is a cool thing. It lets me do peering & various types of transit on a single circuit, & the availability of the route servers is nice. The full mesh of PVCs removes most of the layer 1-2 pain involved with firing up new interactions. The circuit to get there isn't cheap, but it seems worth it based on my experience.
________________________________________________________________________ Jay Ford, Network Engineering Group, Information Technology Services University of Iowa, Iowa City, IA 52242 email: jay-ford@uiowa.edu, phone: 319-335-5555, fax: 319-335-5505
On Mon, 18 Dec 2000, Ben Buxton wrote:
I'm interested in peoples thoughts as to what is happening with the future of ATM NAPs. Are people moving away from them or are they still quite popular? Downsides or upsides? I'm trying to gauge whether they are worthy of investing equipment and hassle into for peering...
From a research university perspective, the AADS NAP is a cool thing. It lets me do peering & various types of transit on a single circuit, & the availability of the route servers is nice. The full mesh of PVCs removes most of the layer 1-2 pain involved with firing up new interactions. The circuit to get there isn't cheap, but it seems worth it based on my experience.
Do you have any expectations with regards to service provided between various peers and the fact that some peers may (depending on your ATM QoS) step onto other peering partners? Cheers, Chris -- Christian Kuhtz <ck@arch.bellsouth.net> -wk, <ck@gnu.org> -hm Sr. Architect, Engineering & Architecture, BellSouth.net, Atlanta, GA, U.S. "I speak for myself only."
I wrote:
From a research university perspective, the AADS NAP is a cool thing. It lets me do peering & various types of transit on a single circuit, & the availability of the route servers is nice. The full mesh of PVCs removes most of the layer 1-2 pain involved with firing up new interactions. The circuit to get there isn't cheap, but it seems worth it based on my experience.
On Mon, 18 Dec 2000, Christian Kuhtz wrote:
Do you have any expectations with regards to service provided between various peers and the fact that some peers may (depending on your ATM QoS) step onto other peering partners?
The short answer is "no". ;^) The longer answer is that I'm aware of the potential chaos & nondeterminism associated with a multi-user contention-based UBR service such as that offered by the AADS NAP. It's certainly an issue, but it doesn't disqualify it from being a useful piece of a connectivity puzzle provided that one utilizes it with eyes open about its characteristics. ________________________________________________________________________ Jay Ford, Network Engineering Group, Information Technology Services University of Iowa, Iowa City, IA 52242 email: jay-ford@uiowa.edu, phone: 319-335-5555, fax: 319-335-5505
On Mon, 18 Dec 2000, Christian Kuhtz wrote:
Do you have any expectations with regards to service provided between various peers and the fact that some peers may (depending on your ATM QoS) step onto other peering partners?
The short answer is "no". ;^)
The longer answer is that I'm aware of the potential chaos & nondeterminism associated with a multi-user contention-based UBR service such as that offered by the AADS NAP. It's certainly an issue, but it doesn't disqualify it from being a useful piece of a connectivity puzzle provided that one utilizes it with eyes open about its characteristics.
I am looking at connecting to more ATM NAPs, but that circuit to the NAP gets pricey. Rather than backhaul a point to point circuit with ATM encapsulation to each NAP, I would like to purchase transport from an ATM service provider. This could be economical especially considering the potential distance between my nearest router and the NAP, and considering the ability to pay per MBit. My question is, what kind of ATM QoS is most appropriate for this kind of service? Even if I'm dealing with UBR at the NAP itself, my path from here to the NAP across my ATM provider's network may require some better QoS than UBR. Any suggestions would be appreciated. Thanks, -- Reid Knuttila Network Engineer Onvoy
Regardless of how good the technologies are, the router vendors have killed ATM as a future nap technology. To use the Cisco example, ATM tops out an OC-12, and has less port density than anything else at that speed (eg 3xGBE in the same slot). While cute, this is useless for an exchange of the future, where Gigabit will be the preferred connection in the sort term, and 10GigE will be the long term solution. Yesterday's 1 meg public peer on 100meg FE is turning into today's 10 meg public peer on 100Meg FE/622Mbps ATM, which is really a step backwards. They need to be on Gigabit soon, as tomorrow they will be the 100meg public peer, and you'll want to upgrade to 10GigE interfaces. If there were OC-48 or OC-192 ATM coming, and/or switches with the density to make that work it would have a future, but alas, that seems to not be in any vendors road map. -- Leo Bicknell - bicknell@ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
Leo Bicknell wrote:
Regardless of how good the technologies are, the router vendors have killed ATM as a future nap technology. To use the Cisco example, ATM tops out an OC-12 ...
If there were OC-48 or OC-192 ATM coming, and/or switches with the density to make that work it would have a future, but alas, that seems to not be in any vendors road map.
My company (Marconi) makes such a switch: http://www.marconi.com/html/solutions/asx4000.htm Non-blocking OC-48c ATM interfaces have been shipping for some time now. -- David
On Mon, Dec 18, 2000 at 07:27:05PM -0500, David Charlap wrote:
If there were OC-48 or OC-192 ATM coming, and/or switches with the density to make that work it would have a future, but alas, that seems to not be in any vendors road map.
My company (Marconi) makes such a switch:
http://www.marconi.com/html/solutions/asx4000.htm
Non-blocking OC-48c ATM interfaces have been shipping for some time now.
Yes for _switches_, yes, and even the ASX4000 doesn't have the density required. At a 40 gig switch, that's only, 16 ports of OC-48 non-blocking, which is well, non-interesting. Consider a successful NAP will have 80-120 connections, and you would want to be able to get them all into the network at OC-48 speed without having to go completely wild on the trunk side. Design that with your switch, tough job. Compare that to an Extreme 6816, which for half the cost can do 192 gige ports in a single chassis, with 128Gbps of bandwidth. Makes the 40 gig ATM switch seem, well, 1/4 as powerful. This is all academic though, neither Cisco nor Juniper make any ATM interfaces over OC-48 for a _ROUTER_, so even if the whole world wanted it, it couldn't be done today. This is the classic problem with ATM, good backbone (switch-to-switch) capabilities and performance, but no edge support. -- Leo Bicknell - bicknell@ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
He said 'density'. 16 ports (is that right?) of OC48 in a single rack is the opposite of 'density'.
My company (Marconi) makes such a switch:
http://www.marconi.com/html/solutions/asx4000.htm
Non-blocking OC-48c ATM interfaces have been shipping for some time now.
-- David
On Tue, Dec 19, 2000 at 08:49:31AM -0500, Alex Rubenstein wrote:
He said 'density'.
16 ports (is that right?) of OC48 in a single rack is the opposite of 'density'.
I think some router vendors should also read that statement very closely. Then s/OC48/OC192/ and read it again. 192 ports of GigE in a half rack is getting close to density. 256 ports of OC-192 in a rack would be great density. (Yes, I am a dreamer.) -- Leo Bicknell - bicknell@ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
Thus spake "David Charlap" <david.charlap@marconi.com>
Leo Bicknell wrote:
Regardless of how good the technologies are, the router vendors have killed ATM as a future nap technology. To use the Cisco example, ATM tops out an OC-12 ...
I thought there weren't commonly available SAR chips for OC48 yet.
If there were OC-48 or OC-192 ATM coming, and/or switches with the density to make that work it would have a future, but alas, that seems to not be in any vendors road map.
My company (Marconi) makes such a switch:
Push as many bits/RU as a typical GE switch and you can reapply for the term "density". ASX4000: Claimed Bandwidth: 40Gbit/s Height: 32 RU BW per RU: 1.25Gbit/s Volume: 14.59 cu.ft. BW per cu.ft.: 2.74Gbit/s Cat6500 (typical GE switch): Claimed Bandwidth: 256Gbit/s Height: 14.4RU BW per RU: 17.78Gbit/s Volume: 4.54 cu.ft. BW per cu.ft.: 56.39Gbit/s I assume other vendors' GE/POS products have a similar density edge over ATM; I was just using the most expedient example.
Non-blocking OC-48c ATM interfaces have been shipping for some time now.
Switching/trunking ATM at OC48/OC192 speeds is relatively trivial. Doing SAR, even on perfectly ordered cells, at those speeds is non-trivial. Packet slicing sucks.
-- David
S | | Stephen Sprunk, K5SSS, CCIE #3723 :|: :|: Network Design Consultant, GSOLE :|||: :|||: New office: RCDN2 in Richardson, TX .:|||||||:..:|||||||:. Email: ssprunk@cisco.com
So, what you're saying is that I should tell all the folks that want to peer via ATM at CMH-IX (which supports 10/100/GE at this time) that they should get bent or get GigE, right? THANK YOU! John Fraizer EnterZone, Inc CMH-IX NAP On Mon, 18 Dec 2000, Leo Bicknell wrote:
Regardless of how good the technologies are, the router vendors have killed ATM as a future nap technology. To use the Cisco example, ATM tops out an OC-12, and has less port density than anything else at that speed (eg 3xGBE in the same slot). While cute, this is useless for an exchange of the future, where Gigabit will be the preferred connection in the sort term, and 10GigE will be the long term solution.
Yesterday's 1 meg public peer on 100meg FE is turning into today's 10 meg public peer on 100Meg FE/622Mbps ATM, which is really a step backwards. They need to be on Gigabit soon, as tomorrow they will be the 100meg public peer, and you'll want to upgrade to 10GigE interfaces.
If there were OC-48 or OC-192 ATM coming, and/or switches with the density to make that work it would have a future, but alas, that seems to not be in any vendors road map.
-- Leo Bicknell - bicknell@ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
On Mon, Dec 18, 2000 at 09:14:53PM -0500, John Fraizer wrote:
So, what you're saying is that I should tell all the folks that want to peer via ATM at CMH-IX (which supports 10/100/GE at this time) that they should get bent or get GigE, right?
THANK YOU!
I had to go google search for this one, and I presume you mean http://www.cmh-ix.net/. If so, it's a good thing the operators made the wise decision to support GigE, otherwise they couldn't possibly expect to meet demand. :-) I do want to applaud the CMH-IX, not because it will be a major exchange point in the future of the Internet, but because it does keep local traffic local. This is a good thing that should be done more. While I won't suggest that CMH-IX should run an ATM (or other) layer two fabric, I know of many smaller ISP's in metro areas who all buy a DS-3 into the same frame relay or atm network so they can "private peer" for the cost of a PVC. This is sort of an "exchange without an exchange". So, should CMH-IX go ATM? Well, given the traffic levels I don't think you could collect enough in fees to pay for an ATM switch. On that basis, ATM is a poor choice. Again, it's a poor choice not because of the technology, but because of the price points and products available. -- Leo Bicknell - bicknell@ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
participants (9)
-
Alex Rubenstein -
Ben Buxton -
Christian Kuhtz -
David Charlap -
Jay Ford -
John Fraizer -
Leo Bicknell -
Reid Knuttila -
Stephen Sprunk