Re: Links on the blink - what will/should mci & sprint do?
| Can a 7500 being fed a single routing view switch IP packts among 6 OC3 AIP | cards running point-to-point AIP-AIP OC3 links to other 7500s at reasonable | speeds, one wonders. If you change "Can" to "Will", this is unclear right now. What apparently is clear (at least to me) after much beating up of the 7500 series beta boxes we got to play with in ICM land and the 7500 series software in the Cisco test lab, is that SprintLink's plan to use 7500s in the four new cities (Atlanta, Kansas City, Cheyenne and Seattle) that will be on-line imminently is reasonable. Firstly the 7500s, even without the smarter technology cards in development by Cisco will currently handle the load requirements of those cities. Secondly since everything currently is done in the RSPs, we have opportunities to try quick and experimental changes to the IOS software at very low-levels that don't exist with the 7000+SSP (but presumably if the proofs-of-concept work really well we can get a port done). Thirdly, there are some very cunning ideas being proposed by various hardware people for making the 7500 do interesting things on the switching front that very probably will get the router to deal with the kinds of loads you're asking about. Finally, the key advantage of the 7500 is the much better ports-per-dollar and ports-per-rack numbers over the 7000s using current technology, and even better numbers as Cisco rolls out the type of channelizing hardware we have been asking for. Specific performance questions wrt the Cisco 7500 series should probably be asked through your favourite Cisco information channels. I'm pretty sure everyone's going to find out how well 7500s work in practice, backbone-wise, Real Soon Now. The big risk, naturally, is that the platform just doesn't hold up in practice, which would force us to do a redesign along the lines of what Gordon has been asking about. Frankly, though, I'm not too worried. Many very smart people have put alot of work into making sure the 7500 doesn't become a complete corporate embarassment for Cisco. Sean.
Sean writes: Thirdly, there are some very cunning ideas being proposed by various hardware people for making the 7500 do interesting things on the switching front that very probably will get the router to deal with the kinds of loads you're asking about. COOK: does this mean that there is some way to get the cisco to behave like the cascade switch I asked about?? If not what in general does it mean? Finally you said if the 7500 doesn't perform well then you might have to do the redesign I asked about.... meaning going to an atm switched backbone with routers at the periphery of the backbone?? ******************************************************************** Gordon Cook, Editor & Publisher Subscript.: Individ-ascii $85 The COOK Report on Internet Non Profit. $150 431 Greenway Ave, Ewing, NJ 08618 Small Corp & Gov't $200 (609) 882-2572 Corporate $350 Internet: cook@cookreport.com Corporate. Site Lic $650 Newly expanded COOK Report Web Pages http://pobox.com/cook/ ********************************************************************
In message <Pine.SUN.3.91.951118235535.11322D-100000@tigger.jvnc.net>, Gordon C ook writes:
Sean writes:
Thirdly, there are some very cunning ideas being proposed by various hardware people for making the 7500 do interesting things on the switching front that very probably will get the router to deal with the kinds of loads you're asking about.
COOK: does this mean that there is some way to get the cisco to behave like the cascade switch I asked about?? If not what in general does it mean?
Finally you said if the 7500 doesn't perform well then you might have to do the redesign I asked about.... meaning going to an atm switched backbone with routers at the periphery of the backbone??
******************************************************************** Gordon Cook, Editor & Publisher Subscript.: Individ-ascii $85 The COOK Report on Internet Non Profit. $150 431 Greenway Ave, Ewing, NJ 08618 Small Corp & Gov't $200 (609) 882-2572 Corporate $350 Internet: cook@cookreport.com Corporate. Site Lic $650 Newly expanded COOK Report Web Pages http://pobox.com/cook/ ********************************************************************
We are bound by non-disclosure to limit what we say about future products of the vendors we have chosen for final evaluation. I don't think it is any big secret that vendors do have new products in the works. IMO- the so called "wall" is not a problem but you need some heftier routers than what is commonly used today. I don't like the idea of going into a layer 2 and letting layer 2 deal with it. You can probably build a working network either way. Equipment cost per port is not our highest priority, it is secondary. Stability and low loss in our backbone is a very high priority. That says a lot about why the technology we ar looking at is different than PSI, Alternet, and Netcom (and maybe a bit different than Sprint too). ANS charges more for a reason. Curtis
oh come on, curtis. don't get holier-than-thou on us. we know quality is important to ANS; quality is important to ALL of us. asserting anything to the contrary is simply insulting to everyone here. EACH of the large networks (including ANS) has done their share of javelin-catching in the past and will do so again at some time in the future. it is, however, useful that we don't all seem to try and catch the same ones at the same time. that's one (surprise) advantage of having different network architectures. once again, genetic diversity is a strength. as for "who's right", that isn't a very interesting discussion because you can't define "right" except in self-referential terms. you can, however, talk about how different designs have different strengths and different distribution of cost functions. some networks make some things easier, and some networks make a different set of things easier. which is better for the application depends on the weights of the terms in the system cost and service value polynomials. (note that there may not be just one of each of those polynomials) So let's not get catty here. a round of "My network's better than your network!" isn't obviously useful. Cheers, and here's to quiet Thanksgiving in everyone's NOC -mo
participants (4)
-
Curtis Villamizar -
Gordon Cook -
Mike O'Dell -
Sean Doran