Bucky,
I too am enjoying this conversation as well. The first thing I think we need to all agree to is that the best technology doesn't always win...unfortunately. That isn't meant to imply that I believe one technology is better for every application. Indeed, IP and ATM have their own inherent benefits. Given what is important to an ISP, IXC, LEC, enterprise, CableCo, etc. they will choose the technology which addressed their given issues. I always have to remind myself of VHS and Beta. We all use VHS today even though there was no question that Beta was a superior technology and provided superior quality. VHS won mainly due to marketing and the clout of all the vendors versus Sony. A de-facto standard emerged...VHS.
That being said let's look at a few things as they relate to IP and ATM.
But you need a connection-oriented mechanism to provide end-to-end prioritization of that traffic. Routers use TCP (layer 4) to provide connection-oriented services. TCP's weakness is that it's built on top of a connectionless protocol (IP).
Hmmm, well you've got me here. I'm not sure why a connection oriented protocol has any advantage over a connectionless protocol when it comes to end to end prioritization. Every router in the network would obviously need to be configured for the same prioritization scheme. The weakness once again in routed networks is speed. Each router would have to check each packet for traffic type (and each traffic type has a configured priority). I think you're eluding to some other weakness. I'd be interested to hear your opinion.
As for how this packet checking is being remedied. Well, there's a thing called Directory Enabled Networks that places decision intelligence into a set of servers. These servers determine priority on an application (layer 7) basis and also provide security, business management, secure VPN's, and other attributes that ATM can not today. These directory enabled networks are also capable of a granularity all the way up to the application layer and can even allow business and conumers to get software updates from their provider at time of release, add users based on groups and automatically recieve group priority and secure server access...et. this list is huge. Directory coupled with layer 3 switching (not TAG - although TAG is an alternative) allows IP nets to do everything an ATM net can do and more albeit an IP net will never have the latency and jitter attributes of an ATM net.
So, this brings me to my original statement - what is critical to each provider will determine what they install. Well, remember ISP's are interested in data and want to get into voice...but fundamentally data is the issue. Here IP nets with ATM backbones make the most sense. In telephony nets, or cable infrastructures real time is critical, here perhaps ATM is a better tool since applications don't vary that much...although they probably will here too - eventually. It should be interesting though to see how this all plays out. BTW: underlying all this is that fact that there appears to be more marketing clout behind IP, and if they make it work ...even to an acceptable level, then IP will rule out....at least on the edge.
RSVP, IP precedence, class-based queueing, etc... will not scale in
ISP/carrier-class networks.
Why do you say this? Again, take a look at the DEN initiative.
These technologies Cisco is pushing require an awful lot of signaling overhead, so much so that they will cause a medium to large-sized network to collapse if implemented throughout the network. This is the problem with handling QoS issues at layer 4, the TCP layer.
The overhead for an IP net is no larger than for an ATM net...in fact one could argue that it's lower since packet sizes can be larger. A compressed header with all the precedence bits and RSVP is 5 bytes in IP....same as a cell. Yet with IP you can run larger frames..and typically one does. No, I'd like to suggest that ATM has the higher overhead which is why it can't scale down to sub T1 speeds well.
The alternative that Ascend is pushing for is to handle QoS issues at the ATM layer. Quality of Service is currently handled with different adaptation layers
Yes, but there are only 5 and even the best switches aren't offering more levels of prioritization than 32 (the Cisco/Strm BPX)...IP with DEN manages up to application layer remember...very very flexible
provide per-VC absolute QoS.
Again this already exists in the STRM products, but I see no difference between this and IP QoS.... Perhaps you could elaborate.
Identifying types of IP traffic is a relatively simple task to conquer--IMO much easier than for TCP/IP to scalably prioritize traffic. And the IETF is working on a standard to do label switching (MPLS). Once this is accomplished, we'll have standards-based ATM switches that can read IP headers.
Yes, but if the precedence information is not in the header than ATM will not be able to prioritize the IP traffic any better than IP. And if the information does exist then IP will do the prioritization and you can run it across you ATM backbone via AAL5 and everyone is happy. The other major problem with this is deployment and management. ATM is management intensive due to its connection oriented requirement...add all the prioritizations and technicians will be configuring the network for years before they can bring it up. This is one of the beauties of DEN - it's automated once you've determined the attributes of certain groups. Then new users can be added to these groups and automatically be configured throughout the net for which servers they can access, prioritization across the infrastructure, what applications they have access to, mail lists, etc... DEN is quite amazing and will change the rules significantly.
Except for some overhead, IP maps quite well to ATM.
Well, we chatted about this. I'll do a bit more homework but I think you'll find that IP headers once compressed (which occurrs end to end) actually has less overhead than ATM (unless you're talking about AAL-CU and voice specifically).
Also, I do not think LU or NT need an IP-centric company like Bay. In my projected scenario, routers will be prevalent at the edge, and hybrid ATM/IP switches will dominate the core. But edge routers are commodity products even now.
Well, this depends on how deep into the core the ATM switches exist. I personally hold the belief that providers realize that charging for bandwidth is a zero sum gain business. The only way they can compete is to charge for services...business management tools, security, secure VPN's, per application QoS and service, etc., etc.,..in this scenario they need network edges that are very smart...to manage all the IP edge devices (read DEN enabled network). We both agree that small business routers are a commoditiy business and even Cisco doesn't make great margins here. Where they make good margins is on their smart edge routers which look more and more like switches everyday. I guess it's a matter of where the line in the sand ends up getting drawn.... I think you're position is that companies like Cisco will get squeezed out by smart ATM backbones and low margin routers. An interesting point of view. Again it will be interesting to see how this all plays out.
Thanks for the discussion Bucky....
OG |
