At 7:14 AM +0000 2/15/03, Larry Letterman wrote: >L3 is usually considered to be wire speed and uses faster >asics... >Routers such as 7200/7500 use older slower hardware to >route... But to answer Nanda's original question, router vs. L3 switch is really a marketing distinction. Yes, _campus_ L3 switches often use different hardware implementations than WAN-oriented routers, but this is a cost engineeering decision. Indeed, cost is more important than speed on SOHO and branch office routers, which require a different set of optimizations.
Are we saying that routers intended to deal with multiple OC-192, like the 12000 or Juniper M40, are slow? The Nortel V15K router (no longer sold) was faster than a 7500, but nobody thought of it as a switch. While it did have multiple forwarding processors, the real difference was that it had a crossbar rather than a shared bus fabric. I worked on the internal design of its successors. I don't think you could go to the IETF or IRTF and find anyone in the ISP world that makes the distinction that "switches" are faster. Multilayer switching has just become, IMNSHO, a marketing term that confuses things. If you really want to look at high speed, consider a true optical (i.e., not optical-electronic-optical) relay. Is that a switch? Especially when it's switching lambdas, it's more of a layer 1 device. Its control, however, may very well be from a layer 3 engine, which runs routing protocols and controls the lambda switch by GMPLS. It isn't useful to say a "L3 switch" is better or worse than a "router". It's necessary, certainly, to identify speeds and feeds, but also to look at other functionality. It's no accident, for example, that a 3550 doesn't have full BGP functionality -- that's a good value engineering decision. Enterprise switches rarely need the advanced QoS functions that a WAN router will. The real difference is between "routing" (more precisely, path determination and setup) and "forwarding". The trend in high-end devices, more and more, is to separate these into different paths. See, for example, the work in the IETF FORCES WG, and know that there are lots of proprietary things in the labs that go much beyond. For SOHO and branch office devices, cost is more an issue than speed. For campus core devices, speed is an important factor, but it can be achieved with parallelism (EtherChannel) and such as well as interface speed. There are a wide range of design choices on the internal fabric, such as main memory in small routers, shared routing memory in Junipers, shared bus as in the 7500, and single or multistage crossbar. > > > >----- Original Message ----- >From: "Nanda" >To: >Sent: Friday, February 14, 2003 4:46 PM >Subject: Layer3 Routers VS Switches [7:63072] > > >> Hi Guys... >> >> We have Layer3 Switches and routers...In what scenario one >would ideally use >> Layer3 switches over routers.. >> Do They have any significant advantage over using >routers.... >> Why do they have layer3 switches when we have routers are >good enough to do >> the job... >> I am confused...I wud appreciate if someone cud clarify. >> >> Thanks in Advance!!!! >> ______________________________________________ >> With Warm Regards... >> Nanda >[EMAIL PROTECTED] Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=63103&t=63072 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
