>Here are some interesting links
>
>
>http://newsroom.cisco.com/dlls/corp_022201.html
>
>http://www.cisco.com/warp/customer/cc/pd/rt/12000/12416/prodlit/itro_ds.htm
>
>http://www.cisco.com/warp/customer/cc/pd/ifaa/oc192/prodlit/cc19_ds.htm
>
>
>The thing is, Juniper's technology is based upon a central bus architecture
shared memory, not shared bus. There is a difference. I don't
have the URL handy, but Cisco has a paper out by the Stanford
University professor who architected the GSR. It does a nice
comparison of the three basic architectures, shared bus, shared
memory, and crossbar. Shared bus runs out of steam at about 2Gbps,
and the 7500 is about the highest end Cisco product that uses it.
Junipers are generally shared memory and GSRs are generally crossbar.
>
>where as the new GSR routers have a processor for each interface card (as
>the Juniper has one central CPU).
>
>I've seen many tests as where the Juniper routers experience a lot of packet
>loss and a decrease in performance and reliability when the node is fully
>configured with a complete set of cards. Each time a card is removed or
>added, there is downtime with traffic interruptions with the Juniper router
>trying to "catch up" with the changes. If you talk with the Engineers at
>Juniper, they will tell you that scalability is their biggest problem with
>their M series routers. You can run with a few, but they won't scale and
>you're not able to run a huge network with them without running into major
>problems.
Historically, Junipers have central problems and Ciscos have
distributed problems. "Nobody is prefect."
>
>The GSRs run the same as if they had only one interface card or if they are
>completely filled with interface cards. Each interface card is managed by
>its own processor so it all runs independent of each other.
True for the forwarding plane, but, in both products, the
control/path determination plane is still single processor. Might or
might not be a limitation, depending on the situation.
>
>Another downfall of the Juniper routers is that an interface card for an
>M160 Juniper router will not work on another Juniper router. For Cisco
>equipment (like most other Cisco products as well...), a card that works
>with one 12416 GSR Router will work on another 12xxx one without any
>problems. Cisco has already tried to keep the interoperability of equipment
>and hardwarre, so that is always nice. You can also upgrade the lower end
>GSRs to the new 12416 hardware also. Juniper's stuff is all individual
>hardware specific.
There's no clean answer here. At some point, maintaining backwards
compatibility and "investment protection" means that you have to
support a less efficient way of doing things that you can do better
with new technology.
>
>It's very much true that Juniper owns 30% of the Enterprise market share and
>that they were the first to come out with the fastest routers, but since
>Cisco has released their new GSR routers...they aren't the only ones
>anymore. And plus it's also true that Cisco was late coming into this
>space, but I think with Cisco's standard of having high quality and control
>procedures, was definitely worth the wait.
Cisco also has the baggage of complexity and support for a large
number of IOS features that are irrelevant to ISP operations. Take
them out, and you simplify the code and make it easier to test. But
you now are supporting a significantly different code base.
JunOS was built only to support ISP operations.
The bottom line is that each customer needs to make a case-by-case
decision. Hopefully, customers that need products in this
performance range have significant clue for making decisions.
I'll freely admit that I haven't measured the latest products versus
one another, but raw best-case forwarding speed is not the only
appropriate selection criterion. There isn't only one.
Other criteria include forwarding rates when packet filtering or
traffic shaping are in effect. Convergence time is yet another
factor -- see my http://draft-berkowitz-bgpcon-00.txt, or, even
better, the multi-authored next draft
http://draft-berkowitz-bgpcon-01.txt, which should be online sometime
next wek.\\el
>I know tons of loyal Cisco
>powered ISPs were waiting for this breakthrough as well.
>
>But I have to admit, Juniper does make some good stuff too (Lots of ex-Cisco
>employees migrated over to Juniper to work there). But I'm partial to Cisco
>and their equipment but I just wanted to help point out that Juniper is no
longer the only one that makes the fastest routers. =)
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
