Router Wars

Chris Holland writes "On the heels of Juniper Networks' recent release of its TX Matrix Platform, Om Malik is giving an interesting overview of current and upcoming battles between protagonists of the Router Game, armed with their Terabit toys."

Comments

Routers

[Anonymous Coward]

Did anyone else think Linksys Routers, hehehehehe *passes out on the couch*

Re:Routers

[w1r3sp33d]

and it's kinda like comparing the millions of compact sedans that everyone owns to the couple of hand built formula one cars. This article is about the formula one cars.

Say what?

[Anonymous Coward]

I'm pretty sure that summary says something meaningful, but heck if I can figure out what it is.

Re:Say what?

[BabyDave]

No-one can tell you what TX Matrix is - you have to RTFA for yourself :)

SUNDAY SUNDAY SUUUUUNNNNDDAAAAAAAYYY

[SuperBanana]

Router Wars

The sound...of 2 teeeeeerrrrrrraaabits...of raw poowweerr.

Watch the Juniper Junker take on the Cisco Crusher this SUNDAY SUNDAY SUNDAY.

Kids's tickets are just five buuuuux!

overly simplistic overview

[jsailor]

if that's not redundant.

This is a large battle, but not one that is won or lost over a few months and not one that is won by comparing simplistic metrics that the press like to use. Software, management, and operations support have always been key in the routing market. Many faster or bigger router companies with unique technologies have gone nowhere. The list is long and depressing. In any case, Cisco has made a dangerous jump ahead by introducing a new operating system that is loosely based on QNX and enables multi-chassis systems. It also enables in-service software upgrades and host of other operations friendly features. Juniper was perceived as having an edge in software, but Cisco will have leapfrogged them if their software delivers (and that's a big if in many people's minds).

Juniper's TX is somewhat handicapped in it's first release (I believe only 2 systems can be linked) and doesn't have a paying customer. Cisco's CRS-1 is limited in interface types in it's first release and has adubious first set of customers. There are many more issues including: weight, power consumption, scalability, support for specific features, handling lawful intercept across a system that large, integration with management systems, etc., etc, etc.

In short, the market is hesitant to purchase either system due to tight CapEx budgets and other pressures. Given the relatively diminutive size of the core router market when compared to edge routing and LAN switching, this a more a battle for prestige than for anything else.

For more info and industry commentary, see:
http://www.lightreading.com/document.asp?doc _id=63 958&site=lightreading
http://www.lightreading.com /document.asp?doc_id=63 916&site=lightreading
http://www.lightreading.com /document.asp?site=test ing&doc_id=63606

Re:overly simplistic overview

[jsailor]

No, they use dual PowerPCs (don't know which model) on the Route Processor card, see:
http://www.cisco.com/en/US/products/ps5763/p roduct s_data_sheet09186a008022d5f1.html

and mostly likely similar chips on the line cards.
Keep in mind that the system makes extensive use of custom ASICs for packet processing, switch fabric, and other functions.

Confused...

[kzinti]

Never known anybody who's even tried a Cisco router. I've been pretty happy with my DeWalt DW625 plunge router - 3 horsepower, electronic variable speed, soft start, and a nice rack-and-pinion depth adjuster. And what is this tera bit everybody is talking about? I've heard of straight bits, v-groove bits, mortising bits, rabbeting bits, cove bits, roundover bits, and tongue-and-groove sets of bits, but never a tera bit. Anybody care to give me the lowdown on this new woodworking equipment?

Re:Confused...

[JustOK]

The terra bit is fer digging holes in da ground. You can use it when you need a bigger pipe.

No winners in the router wars

[Mstrgeek]

Great write up hope you find it educational

http://www.nwfusion.com/columnists/2004/062104noll e.html

Badass new Cisco router's (presentation included)

[topside420]

If you guys are interested in the next-generation of routers from Cisco, be sure to check out this presentation Cisco came to the office and gave regarding all the new bells and whistle of their new line of routers.

Link to power-point presentation (Works great in OO.org): New Cisco Router presentation

I think the coolest thing to come out with these is going to be the GUI router and PIX config. You can see some screenshots of it in the presentation, its mind-boggling and worth drooling over.

These routers also have specialized processors on them for everything they do. They have crypto chips to encrypt/decrypt things, they have DSP cards to decode voice, VPN accelerator chips, chips to process ACLs etc. They also have some badass modules for them including Unity (voice-mail) module for the router itself! A module with full voice-mail capability including a 10GB hard disk to store the messages along with 4+ DSPs on the card to decode the voice traffic going to/from that card. This takes a hell of alot of load of the CPU for more generic tasks.

Anyway, the link again is http://blaze.topside.org/~topside/isr.ppt

I've heard...

[jd]

Santa Clause doesn't consider terabit routers "toys" and so won't be handing them out to good little geeks. He's hoarding them at the North Pole, where he's running the fastest LAN party on Earth...

On a serious note, I very much like the increased competition in the router market. That's good. Nobody gains and everybody loses when there's only one real player in the game.

I would like to see router developers be a little more FOSS-friendly. Hey, I'm not asking Cisco to Open Source IOS - that would never happen - but IOS supports only a small handful of routing protocols and is woefully lacking on QoS support. Whilst Cisco hardware is very likely highly tuned to the protocols they do use, software is software and a module system would be trivial to develop. (This would not be true if Cisco routers were "real" hardware routers, but almost nobody codes in hardware unless they absolutely have to.)

Would it hurt Cisco to support pluggable protocols and QoS algorithms? I can't see how. It would lessen the attractiveness of any competing system that had some feature Cisco themselves didn't support. And if a third-party module proved popular, it would likely be cheaper to buy it than pay a development team to write it from scratch.

This goes for all their competitors, too, of course. Whether it's Juniper, 3Com, or whoever, no company has the time or the resources to develop and maintain code for all the different protocols out there. They can only support the most popular, which may not be the most effective in any given case. (Popular tends to mean a compromise, not just on capability and throughput, but also on maintenance costs, development costs, etc.)

As things stand, Linux has vastly superior packet filtering and QoS support than almost any commercial router on the market. I've not used the *BSDs for a while, but from what I'm hearing, they're comparable or even better in some areas. All this code, all this expertise, all this R&D, and the major manufactuers can't even touch it. That's stupid.

Yes, license issues would probably block any attempt to port Linux modules over. Probably, but not definitely. As in the closed-source modules in Linux argument, dynamic linking can be considered to involve two distinct programs and therefore not in licensing conflict. The BSDs would have no problems at all, regardless.

Why would Cisco care about such code? Or any of the other manufacturers? It's not up to their usual standards, and they wouldn't make money from it.

Because it weakens the argument for moving to someone else. Because third-party modules aren't their problem to support, so they don't need to care about stability. Because anything that cuts R&D costs without cutting the R&D is earning money. Very significant amounts of money.

Because most of the uber-nerds who are involved in network administration are more likely to have a Unix-ish background (and therefore have a mindset geared to extensibility) than a desktop background (where brand-naming has typically won out over technical characteristics).

Finally, because that would allow these router companies to cash in on the media-darling of the moment (Open Source) without compromising on their supposed Intellectual Property rights. Potential gain, no risk of loss, sounds a good exchange to me.

Re:I've heard...

[notarus]

As a few other people have already pointed out, Cisco's IOS supports every ROUTING protocol currently in use out there, assuming you purchased the correct load (not every load supports IS-IS, for example).

Any "fast" router runs almost exclusively in hardware, not in software. Writing hardware code IS hard, because you're trying to do a LOT of often conflicting things (forward packets, filter packets, qos packets). But anytime a CPU gets involved in a packet forward, you're running SLOW. Even OLD cisco routers leap from a measurement of lots of K packets per second to lots and lots of M packets per second just by hardware routing. One of the primary values of advanced Cisco certifications is learning what causes software forwarding-- nothing turns your expensive fast router into a super slow expensive router faster than leaving DCEF for fast switching!

Last, I'm sorry, but neither Linux nor *Bsd's QOS or filtering features are "better" than Cisco's. They are, for the most part, attempts to duplicate the features that are already present in hardware on most Cisco and other routers, or firewalls as appropriate. I am NOT saying that iptables or pf aren't worth anything; I'm just pointing out that you have a very odd view of the state of the art....from 8 years ago.

You DO get a lot more flexibility with iptables or pf at a much lower cost, though. You can do things with iptables on a via 600mhz cpu at 30-40Mb/s that you need $120k worth of combined routers and firewalls to do with more "dedicated" hardware. This isn't always a bad thing for the networking vendors, and it isn't always a good thing for the little PC. You're going to scale much higher on a "real" network after a certain point. You're going to get a lot more flexibility at a lower point with a FOSS router/fw, with the caveat that the scalability is much much harder for an enterprise network.

As with everything else, there is no One True Solution. Pick what works.

mark

Re:Did anyone hear about procket networks ?

[Sideshow Coward]

Cisco bought Procket Networks for the engeering talent that they have attracted with in the last few months, not their product. There is some speculation that Procket would never have a product, but rather being formed as a way for Cisco to recruit engineers from other routing companies. Think about it, the possibilty of having Cisco buy a start up you work for is a nice carrot to have dangled in front of you.

Re:Cost effective.

[duffbeer703]

In Korea, only old people use terabit routers

Re:Argh, what I want...

[PhraudulentOne]

Well, I don't have a lot of experience with SSL offloading (we are an ISP and do webhosting, but we aren't a hosting provider with crazy amounts of SSL-enabled sites), but I met with Cisco a few weeks ago to purchase some new equipment (I don't think I am going to though), and they showed me their 7600 series boxes. One of the blades that you can stick in these is an SSL processor. Click
Here
to check out the link. Here is the summary:

Up to four SSL service modules can be installed in each chassis providing the fastest SSL session setup rates and bulk encrypted throughput in the industry and supporting the highest number of concurrent connections:

3000 connection setups/second per module--10,000 per Chassis fully-populated with SSL modules

300 Mbps bulk encrypted throughput per chassis module--1.2 Gbps per fully-populated with SSL modules

64,000 concurrent client connections--256,000 per chassis fully-populated with SSL modules

So it doesn't look like one blade will do you, but if you stick 4 in there, your rockin'